Notebook 3 of 3
Detectron2 is Facebook AI Research's next generation software system that implements state-of-the-art object detection algorithms. It is a ground-up rewrite of the previous version, Detectron, and it originates from maskrcnn-benchmark. The main feature of the Detectron2 is how they enhanced the training time.
The real power of Detectron2 lies in the HUGE amount of pre-trained models available at the Model Zoo. In addition, the Detectron2 is extendible so a lot of custom configuration can be added.
I chose this model because of the hype that was spread in the world about this new state of the art model and in the competition in particular. I am used to the traditional way of training models, where I need to build a train loop to feed the net with samples and targets, choose a loss function and choose an optimizer. This model has a unique training process which includes data preprocessing and mapping, model configuration and more. The loss functions and optimizer are built in but are configurable.
In this notebook I will demonstrate how to use this models, how to train it and the results of the training.
NOTE: submission scores are found on the first notebook
Installation¶
In [ ]:
!pip install -q -U git+https://github.com/albumentations-team/albumentations
!pip install -q pyyaml==5.1 pycocotools>=2.0.1
!pip install -q detectron2 -f https://dl.fbaipublicfiles.com/detectron2/wheels/cu101/torch1.6/index.html
In [4]:
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import gc
import os
import copy
import cv2
import torch
import torchvision
import pycocotools
import detectron2
import random
import itertools
from tqdm.notebook import tqdm
from glob import glob
from detectron2.config import get_cfg
from detectron2 import model_zoo
from detectron2.engine import DefaultPredictor, DefaultTrainer
from detectron2.utils.visualizer import Visualizer, ColorMode
from detectron2.structures import BoxMode
from detectron2.data import datasets, DatasetCatalog, MetadataCatalog, build_detection_train_loader, build_detection_test_loader
from detectron2.data import transforms as T
from detectron2.data import detection_utils as utils
from detectron2.evaluation import COCOEvaluator, verify_results
from detectron2.modeling import GeneralizedRCNNWithTTA
from detectron2.data.transforms import TransformGen
from detectron2.utils.logger import setup_logger
setup_logger()
from fvcore.transforms.transform import TransformList, Transform, NoOpTransform
from contextlib import contextmanager
SEED = 44
def seed_everything(seed):
random.seed(seed)
os.environ['PYTHONHASHSEED'] = str(seed)
np.random.seed(seed)
torch.manual_seed(seed)
torch.cuda.manual_seed(seed)
torch.backends.cudnn.deterministic = True
torch.backends.cudnn.benchmark = True
seed_everything(SEED)
Configurations¶
In [ ]:
# MODEL_PATH = 'COCO-Detection/faster_rcnn_X_101_32x8d_FPN_3x.yaml'
MODEL_PATH = 'COCO-Detection/retinanet_R_101_FPN_3x.yaml'
TRAIN_IMAGE_PATH = '/content/drive/My Drive/CV/Global Wheat Detection/train'
TEST_IMAGE_PATH = '/content/drive/My Drive/CV/Global Wheat Detection/test'
DATA_PATH = '/content/drive/My Drive/CV/Global Wheat Detection/train.csv'
OUTPUT_PATH ='/content/drive/My Drive/Colab Notebooks/checkpoints/Detectron2'
images_ids = [p.split('/')[-1].split('.')[0] for p in glob(f'{TRAIN_IMAGE_PATH}/*.jpg')]
Import GWD Dataset¶
In [6]:
def read_csv(path) -> pd.DataFrame:
df = pd.read_csv(path)
bboxes = np.stack(df['bbox'].apply(lambda x: np.fromstring(x[1:-1], sep=',')))
for i, column in enumerate(['x_min', 'y_min', 'width', 'height']):
df[column] = bboxes[:,i]
df["x_max"] = df.apply(lambda col: col.x_min + col.width, axis=1)
df["y_max"] = df.apply(lambda col: col.y_min + col.height, axis=1)
df["area"] = df.apply(lambda col: col.width * col.height, axis=1)
df["class"] = 1 # 1 for wheat, 0 for background
df.drop(columns=['source'], inplace=True)
df.drop(columns=['bbox'], inplace=True)
return df
train_df = read_csv(DATA_PATH)
train_df.head()
Out[6]:
| image_id | width | height | x_min | y_min | x_max | y_max | area | class | |
|---|---|---|---|---|---|---|---|---|---|
| 0 | b6ab77fd7 | 56.0 | 36.0 | 834.0 | 222.0 | 890.0 | 258.0 | 2016.0 | 1 |
| 1 | b6ab77fd7 | 130.0 | 58.0 | 226.0 | 548.0 | 356.0 | 606.0 | 7540.0 | 1 |
| 2 | b6ab77fd7 | 74.0 | 160.0 | 377.0 | 504.0 | 451.0 | 664.0 | 11840.0 | 1 |
| 3 | b6ab77fd7 | 109.0 | 107.0 | 834.0 | 95.0 | 943.0 | 202.0 | 11663.0 | 1 |
| 4 | b6ab77fd7 | 124.0 | 117.0 | 26.0 | 144.0 | 150.0 | 261.0 | 14508.0 | 1 |
Custom Augmentation¶
Detectron2 has limited augemntations(Transformation), but they provided the Transform module to make custom augmentations
In [7]:
class CutOut(Transform):
def __init__(self, box_size=25, prob_cutout=0.5):
super().__init__()
self.box_size = box_size
self.prob_cutout = prob_cutout
def apply_image(self, img):
if random.random() > self.prob_cutout:
h, w = img.shape[:2]
num_rand = np.random.randint(10, 20)
for num_cut in range(num_rand):
x_rand, y_rand = random.randint(0, w-self.box_size), random.randint(0, h-self.box_size)
img[x_rand:x_rand+self.box_size, y_rand:y_rand+self.box_size, :] = 0
return np.asarray(img)
def apply_coords(self, coords):
return coords.astype(np.float32)
Data Pre-processing¶
Before jumping into the Training phase preprocessing should be done. first, we need to create a function that will convert our dataset into a format that is used by Detectron2 The format can be reviewed at the tutorial that is provided by FAIR themselves, or can be found at their documentations.
I converted every annotation row to a single record with a list of annotations. You might also notice the polygon that is of the exact same shape as the bounding box. This is required for the image segmentation models in Detectron2.
In [ ]:
IMAGE_SIZE = 1024
def custom_dataset(df, dir_image):
dataset_dicts = []
for img_id, img_name in enumerate(images_ids):
record = {}
image_df = df[df['image_id'] == img_name]
img_path = f'{dir_image}/{img_name}.jpg'
record['file_name'] = img_path
record['image_id'] = img_id
record['height'] = 1024
record['width'] = 1024
objs = []
for _, row in image_df.iterrows():
x_min = int(row.x_min)
y_min = int(row.y_min)
x_max = int(row.x_max)
y_max = int(row.y_max)
poly = [(x_min, y_min), (x_max, y_min),
(x_max, y_max), (x_min, y_max) ]
poly = list(itertools.chain.from_iterable(poly))
obj = {
"bbox": [x_min, y_min, x_max, y_max],
"bbox_mode": BoxMode.XYXY_ABS,
"segmentation": [poly],
"category_id": 0,
"iscrowd" : 0
}
objs.append(obj)
record['annotations'] = objs
dataset_dicts.append(record)
return dataset_dicts
Data Mapper¶
Now, we should implement a mapper function. This function will help us customize the data loader. We this we can add more augmentations and add other configurations for the images.
In [16]:
def custom_mapper(dataset_dict):
# Implement a mapper, similar to the default DatasetMapper, but with your own customizations
dataset_dict = copy.deepcopy(dataset_dict) # it will be modified by code below
image = utils.read_image(dataset_dict["file_name"], format="BGR")
transform_list = [
T.Resize((512,512)),
T.RandomBrightness(0.6, 1.6),
T.RandomContrast(0.6, 3),
T.RandomSaturation(0.1, 2),
T.RandomRotation(angle=[90, 90]),
T.RandomFlip(prob=0.4, horizontal=False, vertical=True),
T.RandomFlip(prob=0.4, horizontal=True, vertical=False),
CutOut()
]
image, transforms = T.apply_transform_gens(transform_list, image)
dataset_dict["image"] = torch.as_tensor(image.transpose(2, 0, 1).astype("float32"))
annos = [
utils.transform_instance_annotations(obj, transforms, image.shape[:2])
for obj in dataset_dict.pop("annotations")
if obj.get("iscrowd", 0) == 0
]
instances = utils.annotations_to_instances(annos, image.shape[:2])
dataset_dict["instances"] = utils.filter_empty_instances(instances)
return dataset_dict
class WheatTrainer(DefaultTrainer):
@classmethod
def build_train_loader(cls, cfg):
return build_detection_train_loader(cfg, mapper=custom_mapper)
Data Registration¶
After the customaztions, I had to regitser the dataset so that the model well recognize it as a legitimate dataset.
In [ ]:
def register_dataset(df, dataset_label='wheat_train', image_dir = TRAIN_IMAGE_PATH):
# Register dataset - if dataset is already registered, give it a new name
try:
DatasetCatalog.register(dataset_label, lambda d=df: custom_dataset(df, image_dir))
MetadataCatalog.get(dataset_label).set(thing_classes = ['wheat'])
except:
# Add random int to dataset name to not run into 'Already registered' error
n = random.randint(1, 1000)
dataset_label = dataset_label + str(n)
DatasetCatalog.register(dataset_label, lambda d=df: custom_dataset(df, image_dir))
MetadataCatalog.get(dataset_label).set(thing_classes = ['wheat'])
return MetadataCatalog.get(dataset_label), dataset_label
metadata, train_dataset = register_dataset(train_df)
Done! Now we can train the model.¶
Detectron2 Configurations¶
Detectron2 has 2 models that can be trained
- FasterRCNN
- RetinaNet
I chose the retina net because Faster RCNN models were used.
RetinaNet¶
RetinaNet is a single, unified network composed of a backbone network and two task-specific subnetworks. The backbone is responsible for computing a conv feature map over an entire input image and is an off-the-self convolution network. The first subnet performs classification on the backbones output; the second subnet performs convolution bounding box regression. The retinanet model uses Focal loss as the classification loss, and SmoothL1 as the box regression loss.
Focal Loss¶
Focal loss is the reshaping of cross entropy loss such that it down-weights the loss assigned to well-classified examples. The novel focal loss focuses training on a sparse set of hard examples and prevents the vast number of easy negatives from overwhelming the detector during training.
After a hours and hours of reviewing the configuration class of the Detectron2's RetinaNet I've finally found how to tune the loss functions parameters, and how to control the learning rates.
SGD¶
Detectron2 uses SGD as the optimizer for all models
more information about the configurations can be found HERE
In [43]:
cfg = get_cfg()
cfg.merge_from_file(model_zoo.get_config_file(MODEL_PATH))
cfg.MODEL.WEIGHTS = model_zoo.get_checkpoint_url(MODEL_PATH) # f'{OUTPUT_PATH}/model_final.pth'
# RETINA
cfg.MODEL.RETINANET.NUM_CLASSES = 1 # Configuring the number of outputs
# RetinaNet Loss parameters
cfg.MODEL.RETINANET.FOCAL_LOSS_GAMMA = 4.0
cfg.MODEL.RETINANET.FOCAL_LOSS_ALPHA = 0.25
cfg.MODEL.RETINANET.SMOOTH_L1_LOSS_BETA = 0.5
cfg.DATASETS.TRAIN = (train_dataset,)
cfg.DATASETS.TEST = ()
cfg.DATALOADER.NUM_WORKERS = 4
cfg.DATALOADER.FILTER_EMPTY_ANNOTATIONS = False # True if u want to exclude "empty" images
cfg.SOLVER.IMS_PER_BATCH = 2
# cfg.SOLVER.LR_SCHEDULER_NAME = 'WarmupCosineLR'
cfg.SOLVER.BASE_LR = 0.00025
cfg.SOLVER.WARMUP_ITERS = 1000
cfg.SOLVER.GAMMA = 0.05
cfg.SOLVER.MAX_ITER = 15000 # 20000 - was used in the competition
cfg.SOLVER.WEIGHT_DECAY = 1e-3
cfg.SOLVER.MOMENTUM = 0.9
cfg.SOLVER.STEPS = (1000,2000,10000,)
cfg.OUTPUT_DIR = OUTPUT_PATH
os.makedirs(cfg.OUTPUT_DIR, exist_ok=True)
trainer = WheatTrainer(cfg)
Loading config /usr/local/lib/python3.6/dist-packages/detectron2/model_zoo/configs/COCO-Detection/../Base-RetinaNet.yaml with yaml.unsafe_load. Your machine may be at risk if the file contains malicious content.
[08/07 15:00:22 d2.engine.defaults]: Model: RetinaNet( (backbone): FPN( (fpn_lateral3): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1)) (fpn_output3): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (fpn_lateral4): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1)) (fpn_output4): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (fpn_lateral5): Conv2d(2048, 256, kernel_size=(1, 1), stride=(1, 1)) (fpn_output5): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (top_block): LastLevelP6P7( (p6): Conv2d(2048, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1)) (p7): Conv2d(256, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1)) ) (bottom_up): ResNet( (stem): BasicStem( (conv1): Conv2d( 3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05) ) ) (res2): Sequential( (0): BottleneckBlock( (shortcut): Conv2d( 64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv1): Conv2d( 64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05) ) (conv2): Conv2d( 64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05) ) (conv3): Conv2d( 64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) ) (1): BottleneckBlock( (conv1): Conv2d( 256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05) ) (conv2): Conv2d( 64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05) ) (conv3): Conv2d( 64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) ) (2): BottleneckBlock( (conv1): Conv2d( 256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05) ) (conv2): Conv2d( 64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05) ) (conv3): Conv2d( 64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) ) ) (res3): Sequential( (0): BottleneckBlock( (shortcut): Conv2d( 256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05) ) (conv1): Conv2d( 256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05) ) (conv2): Conv2d( 128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05) ) (conv3): Conv2d( 128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05) ) ) (1): BottleneckBlock( (conv1): Conv2d( 512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05) ) (conv2): Conv2d( 128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05) ) (conv3): Conv2d( 128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05) ) ) (2): BottleneckBlock( (conv1): Conv2d( 512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05) ) (conv2): Conv2d( 128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05) ) (conv3): Conv2d( 128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05) ) ) (3): BottleneckBlock( (conv1): Conv2d( 512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05) ) (conv2): Conv2d( 128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05) ) (conv3): Conv2d( 128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05) ) ) ) (res4): Sequential( (0): BottleneckBlock( (shortcut): Conv2d( 512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv1): Conv2d( 512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv2): Conv2d( 256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv3): Conv2d( 256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (1): BottleneckBlock( (conv1): Conv2d( 1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv2): Conv2d( 256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv3): Conv2d( 256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (2): BottleneckBlock( (conv1): Conv2d( 1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv2): Conv2d( 256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv3): Conv2d( 256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (3): BottleneckBlock( (conv1): Conv2d( 1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv2): Conv2d( 256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv3): Conv2d( 256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (4): BottleneckBlock( (conv1): Conv2d( 1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv2): Conv2d( 256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv3): Conv2d( 256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (5): BottleneckBlock( (conv1): Conv2d( 1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv2): Conv2d( 256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv3): Conv2d( 256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (6): BottleneckBlock( (conv1): Conv2d( 1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv2): Conv2d( 256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv3): Conv2d( 256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (7): BottleneckBlock( (conv1): Conv2d( 1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv2): Conv2d( 256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv3): Conv2d( 256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (8): BottleneckBlock( (conv1): Conv2d( 1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv2): Conv2d( 256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv3): Conv2d( 256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (9): BottleneckBlock( (conv1): Conv2d( 1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv2): Conv2d( 256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv3): Conv2d( 256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (10): BottleneckBlock( (conv1): Conv2d( 1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv2): Conv2d( 256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv3): Conv2d( 256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (11): BottleneckBlock( (conv1): Conv2d( 1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv2): Conv2d( 256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv3): Conv2d( 256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (12): BottleneckBlock( (conv1): Conv2d( 1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv2): Conv2d( 256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv3): Conv2d( 256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (13): BottleneckBlock( (conv1): Conv2d( 1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv2): Conv2d( 256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv3): Conv2d( 256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (14): BottleneckBlock( (conv1): Conv2d( 1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv2): Conv2d( 256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv3): Conv2d( 256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (15): BottleneckBlock( (conv1): Conv2d( 1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv2): Conv2d( 256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv3): Conv2d( 256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (16): BottleneckBlock( (conv1): Conv2d( 1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv2): Conv2d( 256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv3): Conv2d( 256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (17): BottleneckBlock( (conv1): Conv2d( 1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv2): Conv2d( 256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv3): Conv2d( 256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (18): BottleneckBlock( (conv1): Conv2d( 1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv2): Conv2d( 256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv3): Conv2d( 256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (19): BottleneckBlock( (conv1): Conv2d( 1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv2): Conv2d( 256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv3): Conv2d( 256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (20): BottleneckBlock( (conv1): Conv2d( 1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv2): Conv2d( 256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv3): Conv2d( 256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (21): BottleneckBlock( (conv1): Conv2d( 1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv2): Conv2d( 256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv3): Conv2d( 256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (22): BottleneckBlock( (conv1): Conv2d( 1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv2): Conv2d( 256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv3): Conv2d( 256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) ) (res5): Sequential( (0): BottleneckBlock( (shortcut): Conv2d( 1024, 2048, kernel_size=(1, 1), stride=(2, 2), bias=False (norm): FrozenBatchNorm2d(num_features=2048, eps=1e-05) ) (conv1): Conv2d( 1024, 512, kernel_size=(1, 1), stride=(2, 2), bias=False (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05) ) (conv2): Conv2d( 512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05) ) (conv3): Conv2d( 512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=2048, eps=1e-05) ) ) (1): BottleneckBlock( (conv1): Conv2d( 2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05) ) (conv2): Conv2d( 512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05) ) (conv3): Conv2d( 512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=2048, eps=1e-05) ) ) (2): BottleneckBlock( (conv1): Conv2d( 2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05) ) (conv2): Conv2d( 512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05) ) (conv3): Conv2d( 512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=2048, eps=1e-05) ) ) ) ) ) (head): RetinaNetHead( (cls_subnet): Sequential( (0): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (1): ReLU() (2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (3): ReLU() (4): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (5): ReLU() (6): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (7): ReLU() ) (bbox_subnet): Sequential( (0): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (1): ReLU() (2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (3): ReLU() (4): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (5): ReLU() (6): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (7): ReLU() ) (cls_score): Conv2d(256, 9, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (bbox_pred): Conv2d(256, 36, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) ) (anchor_generator): DefaultAnchorGenerator( (cell_anchors): BufferList() ) ) [08/07 15:01:10 d2.data.common]: Serializing 3422 elements to byte tensors and concatenating them all ... [08/07 15:01:10 d2.data.common]: Serialized dataset takes 9.82 MiB [08/07 15:01:10 d2.data.build]: Using training sampler TrainingSampler
Data visualization¶
In [ ]:
train_data_loader = trainer.build_train_loader(cfg)
data_iter = iter(train_data_loader)
batch = next(data_iter)
In [19]:
rows, cols = 1, 2
plt.figure(figsize=(20,20))
for i, per_image in enumerate(batch[:2]):
plt.subplot(rows, cols, i+1)
# Pytorch tensor is in (C, H, W) format
img = per_image["image"].permute(1, 2, 0).cpu().detach().numpy()
img = utils.convert_image_to_rgb(img, cfg.INPUT.FORMAT)
visualizer = Visualizer(img, metadata=metadata, scale=1.0)
target_fields = per_image["instances"].get_fields()
labels = None
vis = visualizer.overlay_instances(
labels=labels,
boxes=target_fields.get("gt_boxes", None),
masks=target_fields.get("gt_masks", None),
keypoints=target_fields.get("gt_keypoints", None),
)
plt.imshow(vis.get_image()[:, :, ::-1])
In [44]:
trainer.resume_or_load(resume=True)
trainer.train()
[08/07 15:01:12 d2.engine.train_loop]: Starting training from iteration 5000 [08/07 15:01:17 d2.utils.events]: eta: 0:40:01 iter: 5019 total_loss: 0.215 loss_cls: 0.118 loss_box_reg: 0.101 time: 0.2322 data_time: 0.0450 lr: 0.000001 max_mem: 10044M [08/07 15:01:22 d2.utils.events]: eta: 0:38:07 iter: 5039 total_loss: 0.135 loss_cls: 0.072 loss_box_reg: 0.066 time: 0.2270 data_time: 0.0133 lr: 0.000001 max_mem: 10044M [08/07 15:01:27 d2.utils.events]: eta: 0:38:03 iter: 5059 total_loss: 0.129 loss_cls: 0.065 loss_box_reg: 0.065 time: 0.2278 data_time: 0.0136 lr: 0.000001 max_mem: 10044M [08/07 15:01:31 d2.utils.events]: eta: 0:38:01 iter: 5079 total_loss: 0.143 loss_cls: 0.074 loss_box_reg: 0.068 time: 0.2271 data_time: 0.0117 lr: 0.000001 max_mem: 10044M [08/07 15:01:36 d2.utils.events]: eta: 0:37:59 iter: 5099 total_loss: 0.150 loss_cls: 0.077 loss_box_reg: 0.072 time: 0.2279 data_time: 0.0115 lr: 0.000001 max_mem: 10044M [08/07 15:01:40 d2.utils.events]: eta: 0:37:54 iter: 5119 total_loss: 0.126 loss_cls: 0.067 loss_box_reg: 0.061 time: 0.2271 data_time: 0.0095 lr: 0.000001 max_mem: 10044M [08/07 15:01:45 d2.utils.events]: eta: 0:37:50 iter: 5139 total_loss: 0.139 loss_cls: 0.071 loss_box_reg: 0.068 time: 0.2274 data_time: 0.0133 lr: 0.000001 max_mem: 10044M [08/07 15:01:49 d2.utils.events]: eta: 0:37:49 iter: 5159 total_loss: 0.146 loss_cls: 0.076 loss_box_reg: 0.070 time: 0.2283 data_time: 0.0126 lr: 0.000001 max_mem: 10044M [08/07 15:01:54 d2.utils.events]: eta: 0:37:30 iter: 5179 total_loss: 0.137 loss_cls: 0.073 loss_box_reg: 0.066 time: 0.2275 data_time: 0.0105 lr: 0.000001 max_mem: 10044M [08/07 15:01:59 d2.utils.events]: eta: 0:37:36 iter: 5199 total_loss: 0.134 loss_cls: 0.067 loss_box_reg: 0.066 time: 0.2278 data_time: 0.0126 lr: 0.000001 max_mem: 10044M [08/07 15:02:03 d2.utils.events]: eta: 0:37:21 iter: 5219 total_loss: 0.133 loss_cls: 0.071 loss_box_reg: 0.064 time: 0.2273 data_time: 0.0123 lr: 0.000001 max_mem: 10044M [08/07 15:02:08 d2.utils.events]: eta: 0:37:13 iter: 5239 total_loss: 0.142 loss_cls: 0.072 loss_box_reg: 0.072 time: 0.2273 data_time: 0.0134 lr: 0.000001 max_mem: 10044M [08/07 15:02:12 d2.utils.events]: eta: 0:37:08 iter: 5259 total_loss: 0.141 loss_cls: 0.071 loss_box_reg: 0.069 time: 0.2273 data_time: 0.0109 lr: 0.000001 max_mem: 10044M [08/07 15:02:17 d2.utils.events]: eta: 0:37:03 iter: 5279 total_loss: 0.138 loss_cls: 0.072 loss_box_reg: 0.068 time: 0.2270 data_time: 0.0131 lr: 0.000001 max_mem: 10044M [08/07 15:02:21 d2.utils.events]: eta: 0:36:58 iter: 5299 total_loss: 0.136 loss_cls: 0.066 loss_box_reg: 0.069 time: 0.2270 data_time: 0.0129 lr: 0.000001 max_mem: 10044M [08/07 15:02:26 d2.utils.events]: eta: 0:36:54 iter: 5319 total_loss: 0.135 loss_cls: 0.070 loss_box_reg: 0.066 time: 0.2267 data_time: 0.0129 lr: 0.000001 max_mem: 10044M [08/07 15:02:30 d2.utils.events]: eta: 0:36:46 iter: 5339 total_loss: 0.135 loss_cls: 0.070 loss_box_reg: 0.059 time: 0.2263 data_time: 0.0129 lr: 0.000001 max_mem: 10044M [08/07 15:02:35 d2.utils.events]: eta: 0:36:40 iter: 5359 total_loss: 0.149 loss_cls: 0.075 loss_box_reg: 0.073 time: 0.2261 data_time: 0.0135 lr: 0.000001 max_mem: 10044M [08/07 15:02:39 d2.utils.events]: eta: 0:36:36 iter: 5379 total_loss: 0.136 loss_cls: 0.068 loss_box_reg: 0.069 time: 0.2262 data_time: 0.0110 lr: 0.000001 max_mem: 10044M [08/07 15:02:44 d2.utils.events]: eta: 0:36:33 iter: 5399 total_loss: 0.147 loss_cls: 0.072 loss_box_reg: 0.074 time: 0.2264 data_time: 0.0122 lr: 0.000001 max_mem: 10044M [08/07 15:02:48 d2.utils.events]: eta: 0:36:29 iter: 5419 total_loss: 0.138 loss_cls: 0.068 loss_box_reg: 0.067 time: 0.2266 data_time: 0.0158 lr: 0.000001 max_mem: 10044M [08/07 15:02:53 d2.utils.events]: eta: 0:36:22 iter: 5439 total_loss: 0.137 loss_cls: 0.069 loss_box_reg: 0.067 time: 0.2267 data_time: 0.0115 lr: 0.000001 max_mem: 10044M [08/07 15:02:58 d2.utils.events]: eta: 0:36:20 iter: 5459 total_loss: 0.140 loss_cls: 0.073 loss_box_reg: 0.066 time: 0.2268 data_time: 0.0126 lr: 0.000001 max_mem: 10044M [08/07 15:03:02 d2.utils.events]: eta: 0:36:19 iter: 5479 total_loss: 0.139 loss_cls: 0.075 loss_box_reg: 0.071 time: 0.2270 data_time: 0.0121 lr: 0.000001 max_mem: 10044M [08/07 15:03:07 d2.utils.events]: eta: 0:36:12 iter: 5499 total_loss: 0.141 loss_cls: 0.075 loss_box_reg: 0.065 time: 0.2268 data_time: 0.0133 lr: 0.000001 max_mem: 10044M [08/07 15:03:11 d2.utils.events]: eta: 0:36:06 iter: 5519 total_loss: 0.128 loss_cls: 0.067 loss_box_reg: 0.061 time: 0.2267 data_time: 0.0131 lr: 0.000001 max_mem: 10044M [08/07 15:03:16 d2.utils.events]: eta: 0:36:01 iter: 5539 total_loss: 0.151 loss_cls: 0.079 loss_box_reg: 0.075 time: 0.2268 data_time: 0.0139 lr: 0.000001 max_mem: 10044M [08/07 15:03:20 d2.utils.events]: eta: 0:35:57 iter: 5559 total_loss: 0.145 loss_cls: 0.076 loss_box_reg: 0.068 time: 0.2270 data_time: 0.0136 lr: 0.000001 max_mem: 10044M [08/07 15:03:25 d2.utils.events]: eta: 0:35:52 iter: 5579 total_loss: 0.130 loss_cls: 0.068 loss_box_reg: 0.065 time: 0.2271 data_time: 0.0120 lr: 0.000001 max_mem: 10044M [08/07 15:03:30 d2.utils.events]: eta: 0:35:48 iter: 5599 total_loss: 0.126 loss_cls: 0.063 loss_box_reg: 0.062 time: 0.2271 data_time: 0.0118 lr: 0.000001 max_mem: 10044M [08/07 15:03:34 d2.utils.events]: eta: 0:35:41 iter: 5619 total_loss: 0.127 loss_cls: 0.066 loss_box_reg: 0.064 time: 0.2267 data_time: 0.0134 lr: 0.000001 max_mem: 10044M [08/07 15:03:38 d2.utils.events]: eta: 0:35:37 iter: 5639 total_loss: 0.143 loss_cls: 0.075 loss_box_reg: 0.070 time: 0.2265 data_time: 0.0114 lr: 0.000001 max_mem: 10044M [08/07 15:03:43 d2.utils.events]: eta: 0:35:20 iter: 5659 total_loss: 0.146 loss_cls: 0.076 loss_box_reg: 0.069 time: 0.2264 data_time: 0.0143 lr: 0.000001 max_mem: 10044M [08/07 15:03:47 d2.utils.events]: eta: 0:35:16 iter: 5679 total_loss: 0.133 loss_cls: 0.070 loss_box_reg: 0.062 time: 0.2262 data_time: 0.0115 lr: 0.000001 max_mem: 10044M [08/07 15:03:52 d2.utils.events]: eta: 0:35:07 iter: 5699 total_loss: 0.136 loss_cls: 0.071 loss_box_reg: 0.065 time: 0.2263 data_time: 0.0138 lr: 0.000001 max_mem: 10044M [08/07 15:03:56 d2.utils.events]: eta: 0:35:02 iter: 5719 total_loss: 0.145 loss_cls: 0.071 loss_box_reg: 0.069 time: 0.2263 data_time: 0.0127 lr: 0.000001 max_mem: 10044M [08/07 15:04:01 d2.utils.events]: eta: 0:34:55 iter: 5739 total_loss: 0.155 loss_cls: 0.079 loss_box_reg: 0.076 time: 0.2260 data_time: 0.0127 lr: 0.000001 max_mem: 10044M [08/07 15:04:05 d2.utils.events]: eta: 0:34:49 iter: 5759 total_loss: 0.143 loss_cls: 0.076 loss_box_reg: 0.069 time: 0.2258 data_time: 0.0096 lr: 0.000001 max_mem: 10044M [08/07 15:04:10 d2.utils.events]: eta: 0:34:44 iter: 5779 total_loss: 0.144 loss_cls: 0.074 loss_box_reg: 0.070 time: 0.2257 data_time: 0.0119 lr: 0.000001 max_mem: 10044M [08/07 15:04:14 d2.utils.events]: eta: 0:34:39 iter: 5799 total_loss: 0.127 loss_cls: 0.066 loss_box_reg: 0.060 time: 0.2256 data_time: 0.0101 lr: 0.000001 max_mem: 10044M [08/07 15:04:18 d2.utils.events]: eta: 0:34:34 iter: 5819 total_loss: 0.143 loss_cls: 0.076 loss_box_reg: 0.069 time: 0.2255 data_time: 0.0094 lr: 0.000001 max_mem: 10044M [08/07 15:04:23 d2.utils.events]: eta: 0:34:30 iter: 5839 total_loss: 0.131 loss_cls: 0.067 loss_box_reg: 0.065 time: 0.2256 data_time: 0.0123 lr: 0.000001 max_mem: 10044M [08/07 15:04:27 d2.utils.events]: eta: 0:34:25 iter: 5859 total_loss: 0.144 loss_cls: 0.075 loss_box_reg: 0.071 time: 0.2255 data_time: 0.0123 lr: 0.000001 max_mem: 10044M [08/07 15:04:32 d2.utils.events]: eta: 0:34:20 iter: 5879 total_loss: 0.143 loss_cls: 0.074 loss_box_reg: 0.070 time: 0.2254 data_time: 0.0121 lr: 0.000001 max_mem: 10044M [08/07 15:04:37 d2.utils.events]: eta: 0:34:15 iter: 5899 total_loss: 0.137 loss_cls: 0.067 loss_box_reg: 0.067 time: 0.2255 data_time: 0.0133 lr: 0.000001 max_mem: 10044M [08/07 15:04:41 d2.utils.events]: eta: 0:34:12 iter: 5919 total_loss: 0.141 loss_cls: 0.073 loss_box_reg: 0.069 time: 0.2258 data_time: 0.0131 lr: 0.000001 max_mem: 10044M [08/07 15:04:46 d2.utils.events]: eta: 0:34:09 iter: 5939 total_loss: 0.123 loss_cls: 0.063 loss_box_reg: 0.060 time: 0.2261 data_time: 0.0149 lr: 0.000001 max_mem: 10044M [08/07 15:04:51 d2.utils.events]: eta: 0:34:07 iter: 5959 total_loss: 0.138 loss_cls: 0.074 loss_box_reg: 0.069 time: 0.2261 data_time: 0.0137 lr: 0.000001 max_mem: 10044M [08/07 15:04:55 d2.utils.events]: eta: 0:34:03 iter: 5979 total_loss: 0.137 loss_cls: 0.074 loss_box_reg: 0.066 time: 0.2262 data_time: 0.0121 lr: 0.000001 max_mem: 10044M [08/07 15:05:00 d2.utils.events]: eta: 0:33:59 iter: 5999 total_loss: 0.158 loss_cls: 0.080 loss_box_reg: 0.077 time: 0.2264 data_time: 0.0132 lr: 0.000001 max_mem: 10044M [08/07 15:05:05 d2.utils.events]: eta: 0:33:55 iter: 6019 total_loss: 0.144 loss_cls: 0.075 loss_box_reg: 0.068 time: 0.2266 data_time: 0.0135 lr: 0.000001 max_mem: 10044M [08/07 15:05:09 d2.utils.events]: eta: 0:33:58 iter: 6039 total_loss: 0.119 loss_cls: 0.064 loss_box_reg: 0.057 time: 0.2268 data_time: 0.0131 lr: 0.000001 max_mem: 10044M [08/07 15:05:14 d2.utils.events]: eta: 0:33:51 iter: 6059 total_loss: 0.137 loss_cls: 0.068 loss_box_reg: 0.069 time: 0.2267 data_time: 0.0136 lr: 0.000001 max_mem: 10044M [08/07 15:05:19 d2.utils.events]: eta: 0:33:49 iter: 6079 total_loss: 0.139 loss_cls: 0.071 loss_box_reg: 0.066 time: 0.2268 data_time: 0.0140 lr: 0.000001 max_mem: 10044M [08/07 15:05:23 d2.utils.events]: eta: 0:33:42 iter: 6099 total_loss: 0.127 loss_cls: 0.065 loss_box_reg: 0.063 time: 0.2269 data_time: 0.0152 lr: 0.000001 max_mem: 10044M [08/07 15:05:28 d2.utils.events]: eta: 0:33:40 iter: 6119 total_loss: 0.133 loss_cls: 0.072 loss_box_reg: 0.062 time: 0.2274 data_time: 0.0149 lr: 0.000001 max_mem: 10044M [08/07 15:05:33 d2.utils.events]: eta: 0:33:38 iter: 6139 total_loss: 0.137 loss_cls: 0.071 loss_box_reg: 0.068 time: 0.2275 data_time: 0.0127 lr: 0.000001 max_mem: 10044M [08/07 15:05:38 d2.utils.events]: eta: 0:33:35 iter: 6159 total_loss: 0.131 loss_cls: 0.065 loss_box_reg: 0.062 time: 0.2276 data_time: 0.0134 lr: 0.000001 max_mem: 10044M [08/07 15:05:42 d2.utils.events]: eta: 0:33:32 iter: 6179 total_loss: 0.143 loss_cls: 0.072 loss_box_reg: 0.071 time: 0.2277 data_time: 0.0150 lr: 0.000001 max_mem: 10044M [08/07 15:05:47 d2.utils.events]: eta: 0:33:29 iter: 6199 total_loss: 0.141 loss_cls: 0.073 loss_box_reg: 0.066 time: 0.2278 data_time: 0.0146 lr: 0.000001 max_mem: 10044M [08/07 15:05:52 d2.utils.events]: eta: 0:33:27 iter: 6219 total_loss: 0.143 loss_cls: 0.073 loss_box_reg: 0.067 time: 0.2280 data_time: 0.0127 lr: 0.000001 max_mem: 10044M [08/07 15:05:57 d2.utils.events]: eta: 0:33:25 iter: 6239 total_loss: 0.132 loss_cls: 0.070 loss_box_reg: 0.063 time: 0.2281 data_time: 0.0140 lr: 0.000001 max_mem: 10044M [08/07 15:06:01 d2.utils.events]: eta: 0:33:24 iter: 6259 total_loss: 0.147 loss_cls: 0.079 loss_box_reg: 0.070 time: 0.2282 data_time: 0.0125 lr: 0.000001 max_mem: 10044M [08/07 15:06:06 d2.utils.events]: eta: 0:33:18 iter: 6279 total_loss: 0.135 loss_cls: 0.072 loss_box_reg: 0.065 time: 0.2282 data_time: 0.0111 lr: 0.000001 max_mem: 10044M [08/07 15:06:10 d2.utils.events]: eta: 0:33:13 iter: 6299 total_loss: 0.143 loss_cls: 0.073 loss_box_reg: 0.068 time: 0.2281 data_time: 0.0127 lr: 0.000001 max_mem: 10044M [08/07 15:06:15 d2.utils.events]: eta: 0:33:13 iter: 6319 total_loss: 0.137 loss_cls: 0.068 loss_box_reg: 0.070 time: 0.2282 data_time: 0.0133 lr: 0.000001 max_mem: 10044M [08/07 15:06:20 d2.utils.events]: eta: 0:33:09 iter: 6339 total_loss: 0.145 loss_cls: 0.077 loss_box_reg: 0.066 time: 0.2282 data_time: 0.0144 lr: 0.000001 max_mem: 10044M [08/07 15:06:24 d2.utils.events]: eta: 0:33:09 iter: 6359 total_loss: 0.141 loss_cls: 0.075 loss_box_reg: 0.068 time: 0.2283 data_time: 0.0145 lr: 0.000001 max_mem: 10044M [08/07 15:06:29 d2.utils.events]: eta: 0:33:05 iter: 6379 total_loss: 0.141 loss_cls: 0.070 loss_box_reg: 0.072 time: 0.2285 data_time: 0.0169 lr: 0.000001 max_mem: 10044M [08/07 15:06:34 d2.utils.events]: eta: 0:33:01 iter: 6399 total_loss: 0.135 loss_cls: 0.067 loss_box_reg: 0.067 time: 0.2287 data_time: 0.0153 lr: 0.000001 max_mem: 10044M [08/07 15:06:39 d2.utils.events]: eta: 0:32:58 iter: 6419 total_loss: 0.136 loss_cls: 0.071 loss_box_reg: 0.065 time: 0.2289 data_time: 0.0144 lr: 0.000001 max_mem: 10044M [08/07 15:06:44 d2.utils.events]: eta: 0:32:55 iter: 6439 total_loss: 0.144 loss_cls: 0.071 loss_box_reg: 0.070 time: 0.2290 data_time: 0.0136 lr: 0.000001 max_mem: 10044M [08/07 15:06:49 d2.utils.events]: eta: 0:32:50 iter: 6459 total_loss: 0.130 loss_cls: 0.066 loss_box_reg: 0.066 time: 0.2291 data_time: 0.0136 lr: 0.000001 max_mem: 10044M [08/07 15:06:53 d2.utils.events]: eta: 0:32:44 iter: 6479 total_loss: 0.145 loss_cls: 0.074 loss_box_reg: 0.070 time: 0.2292 data_time: 0.0133 lr: 0.000001 max_mem: 10044M [08/07 15:06:58 d2.utils.events]: eta: 0:32:42 iter: 6499 total_loss: 0.133 loss_cls: 0.068 loss_box_reg: 0.068 time: 0.2293 data_time: 0.0168 lr: 0.000001 max_mem: 10044M [08/07 15:07:03 d2.utils.events]: eta: 0:32:41 iter: 6519 total_loss: 0.137 loss_cls: 0.071 loss_box_reg: 0.069 time: 0.2293 data_time: 0.0118 lr: 0.000001 max_mem: 10044M [08/07 15:07:07 d2.utils.events]: eta: 0:32:37 iter: 6539 total_loss: 0.123 loss_cls: 0.065 loss_box_reg: 0.058 time: 0.2294 data_time: 0.0156 lr: 0.000001 max_mem: 10044M [08/07 15:07:12 d2.utils.events]: eta: 0:32:34 iter: 6559 total_loss: 0.137 loss_cls: 0.069 loss_box_reg: 0.068 time: 0.2296 data_time: 0.0144 lr: 0.000001 max_mem: 10044M [08/07 15:07:17 d2.utils.events]: eta: 0:32:30 iter: 6579 total_loss: 0.145 loss_cls: 0.075 loss_box_reg: 0.068 time: 0.2296 data_time: 0.0153 lr: 0.000001 max_mem: 10044M [08/07 15:07:22 d2.utils.events]: eta: 0:32:25 iter: 6599 total_loss: 0.128 loss_cls: 0.066 loss_box_reg: 0.065 time: 0.2297 data_time: 0.0140 lr: 0.000001 max_mem: 10044M [08/07 15:07:26 d2.utils.events]: eta: 0:32:24 iter: 6619 total_loss: 0.150 loss_cls: 0.081 loss_box_reg: 0.072 time: 0.2297 data_time: 0.0130 lr: 0.000001 max_mem: 10044M [08/07 15:07:31 d2.utils.events]: eta: 0:32:20 iter: 6639 total_loss: 0.149 loss_cls: 0.079 loss_box_reg: 0.068 time: 0.2298 data_time: 0.0125 lr: 0.000001 max_mem: 10044M [08/07 15:07:36 d2.utils.events]: eta: 0:32:15 iter: 6659 total_loss: 0.141 loss_cls: 0.069 loss_box_reg: 0.069 time: 0.2298 data_time: 0.0106 lr: 0.000001 max_mem: 10044M [08/07 15:07:40 d2.utils.events]: eta: 0:32:12 iter: 6679 total_loss: 0.126 loss_cls: 0.069 loss_box_reg: 0.057 time: 0.2298 data_time: 0.0153 lr: 0.000001 max_mem: 10044M [08/07 15:07:45 d2.utils.events]: eta: 0:32:08 iter: 6699 total_loss: 0.116 loss_cls: 0.059 loss_box_reg: 0.057 time: 0.2297 data_time: 0.0134 lr: 0.000001 max_mem: 10044M [08/07 15:07:50 d2.utils.events]: eta: 0:32:05 iter: 6719 total_loss: 0.142 loss_cls: 0.076 loss_box_reg: 0.067 time: 0.2298 data_time: 0.0136 lr: 0.000001 max_mem: 10044M [08/07 15:07:54 d2.utils.events]: eta: 0:32:05 iter: 6739 total_loss: 0.149 loss_cls: 0.079 loss_box_reg: 0.069 time: 0.2299 data_time: 0.0105 lr: 0.000001 max_mem: 10044M [08/07 15:07:59 d2.utils.events]: eta: 0:32:02 iter: 6759 total_loss: 0.150 loss_cls: 0.073 loss_box_reg: 0.072 time: 0.2300 data_time: 0.0128 lr: 0.000001 max_mem: 10044M [08/07 15:08:04 d2.utils.events]: eta: 0:31:59 iter: 6779 total_loss: 0.140 loss_cls: 0.070 loss_box_reg: 0.074 time: 0.2301 data_time: 0.0139 lr: 0.000001 max_mem: 10044M [08/07 15:08:09 d2.utils.events]: eta: 0:31:56 iter: 6799 total_loss: 0.121 loss_cls: 0.065 loss_box_reg: 0.056 time: 0.2302 data_time: 0.0149 lr: 0.000001 max_mem: 10044M [08/07 15:08:13 d2.utils.events]: eta: 0:31:55 iter: 6819 total_loss: 0.138 loss_cls: 0.073 loss_box_reg: 0.066 time: 0.2302 data_time: 0.0165 lr: 0.000001 max_mem: 10044M [08/07 15:08:18 d2.utils.events]: eta: 0:31:51 iter: 6839 total_loss: 0.126 loss_cls: 0.066 loss_box_reg: 0.059 time: 0.2304 data_time: 0.0156 lr: 0.000001 max_mem: 10044M [08/07 15:08:23 d2.utils.events]: eta: 0:31:51 iter: 6859 total_loss: 0.146 loss_cls: 0.075 loss_box_reg: 0.072 time: 0.2305 data_time: 0.0138 lr: 0.000001 max_mem: 10044M [08/07 15:08:28 d2.utils.events]: eta: 0:31:47 iter: 6879 total_loss: 0.131 loss_cls: 0.068 loss_box_reg: 0.067 time: 0.2305 data_time: 0.0148 lr: 0.000001 max_mem: 10044M [08/07 15:08:32 d2.utils.events]: eta: 0:31:43 iter: 6899 total_loss: 0.124 loss_cls: 0.065 loss_box_reg: 0.060 time: 0.2305 data_time: 0.0140 lr: 0.000001 max_mem: 10044M [08/07 15:08:37 d2.utils.events]: eta: 0:31:38 iter: 6919 total_loss: 0.137 loss_cls: 0.069 loss_box_reg: 0.066 time: 0.2305 data_time: 0.0173 lr: 0.000001 max_mem: 10044M [08/07 15:08:42 d2.utils.events]: eta: 0:31:32 iter: 6939 total_loss: 0.153 loss_cls: 0.080 loss_box_reg: 0.076 time: 0.2306 data_time: 0.0138 lr: 0.000001 max_mem: 10044M [08/07 15:08:46 d2.utils.events]: eta: 0:31:23 iter: 6959 total_loss: 0.145 loss_cls: 0.074 loss_box_reg: 0.070 time: 0.2305 data_time: 0.0150 lr: 0.000001 max_mem: 10044M [08/07 15:08:51 d2.utils.events]: eta: 0:31:22 iter: 6979 total_loss: 0.142 loss_cls: 0.075 loss_box_reg: 0.068 time: 0.2306 data_time: 0.0162 lr: 0.000001 max_mem: 10044M [08/07 15:08:56 d2.utils.events]: eta: 0:31:18 iter: 6999 total_loss: 0.134 loss_cls: 0.070 loss_box_reg: 0.064 time: 0.2307 data_time: 0.0141 lr: 0.000001 max_mem: 10044M [08/07 15:09:00 d2.utils.events]: eta: 0:31:11 iter: 7019 total_loss: 0.133 loss_cls: 0.067 loss_box_reg: 0.065 time: 0.2306 data_time: 0.0147 lr: 0.000001 max_mem: 10044M [08/07 15:09:05 d2.utils.events]: eta: 0:31:06 iter: 7039 total_loss: 0.141 loss_cls: 0.077 loss_box_reg: 0.066 time: 0.2307 data_time: 0.0124 lr: 0.000001 max_mem: 10044M [08/07 15:09:10 d2.utils.events]: eta: 0:31:03 iter: 7059 total_loss: 0.142 loss_cls: 0.073 loss_box_reg: 0.068 time: 0.2307 data_time: 0.0139 lr: 0.000001 max_mem: 10044M [08/07 15:09:14 d2.utils.events]: eta: 0:30:54 iter: 7079 total_loss: 0.131 loss_cls: 0.069 loss_box_reg: 0.065 time: 0.2307 data_time: 0.0113 lr: 0.000001 max_mem: 10044M [08/07 15:09:19 d2.utils.events]: eta: 0:30:49 iter: 7099 total_loss: 0.129 loss_cls: 0.067 loss_box_reg: 0.062 time: 0.2307 data_time: 0.0113 lr: 0.000001 max_mem: 10044M [08/07 15:09:24 d2.utils.events]: eta: 0:30:43 iter: 7119 total_loss: 0.155 loss_cls: 0.077 loss_box_reg: 0.076 time: 0.2307 data_time: 0.0112 lr: 0.000001 max_mem: 10044M [08/07 15:09:28 d2.utils.events]: eta: 0:30:36 iter: 7139 total_loss: 0.140 loss_cls: 0.071 loss_box_reg: 0.068 time: 0.2306 data_time: 0.0128 lr: 0.000001 max_mem: 10044M [08/07 15:09:33 d2.utils.events]: eta: 0:30:29 iter: 7159 total_loss: 0.144 loss_cls: 0.073 loss_box_reg: 0.072 time: 0.2305 data_time: 0.0139 lr: 0.000001 max_mem: 10044M [08/07 15:09:37 d2.utils.events]: eta: 0:30:24 iter: 7179 total_loss: 0.140 loss_cls: 0.070 loss_box_reg: 0.067 time: 0.2305 data_time: 0.0120 lr: 0.000001 max_mem: 10044M [08/07 15:09:42 d2.utils.events]: eta: 0:30:19 iter: 7199 total_loss: 0.134 loss_cls: 0.070 loss_box_reg: 0.064 time: 0.2305 data_time: 0.0126 lr: 0.000001 max_mem: 10044M [08/07 15:09:46 d2.utils.events]: eta: 0:30:15 iter: 7219 total_loss: 0.140 loss_cls: 0.070 loss_box_reg: 0.070 time: 0.2305 data_time: 0.0133 lr: 0.000001 max_mem: 10044M [08/07 15:09:51 d2.utils.events]: eta: 0:30:09 iter: 7239 total_loss: 0.146 loss_cls: 0.072 loss_box_reg: 0.074 time: 0.2305 data_time: 0.0112 lr: 0.000001 max_mem: 10044M [08/07 15:09:56 d2.utils.events]: eta: 0:30:02 iter: 7259 total_loss: 0.135 loss_cls: 0.067 loss_box_reg: 0.067 time: 0.2304 data_time: 0.0133 lr: 0.000001 max_mem: 10044M [08/07 15:10:00 d2.utils.events]: eta: 0:29:59 iter: 7279 total_loss: 0.143 loss_cls: 0.075 loss_box_reg: 0.068 time: 0.2305 data_time: 0.0158 lr: 0.000001 max_mem: 10044M [08/07 15:10:05 d2.utils.events]: eta: 0:29:55 iter: 7299 total_loss: 0.139 loss_cls: 0.074 loss_box_reg: 0.068 time: 0.2305 data_time: 0.0129 lr: 0.000001 max_mem: 10044M [08/07 15:10:10 d2.utils.events]: eta: 0:29:51 iter: 7319 total_loss: 0.138 loss_cls: 0.070 loss_box_reg: 0.067 time: 0.2305 data_time: 0.0129 lr: 0.000001 max_mem: 10044M [08/07 15:10:14 d2.utils.events]: eta: 0:29:47 iter: 7339 total_loss: 0.138 loss_cls: 0.070 loss_box_reg: 0.067 time: 0.2305 data_time: 0.0140 lr: 0.000001 max_mem: 10044M [08/07 15:10:19 d2.utils.events]: eta: 0:29:39 iter: 7359 total_loss: 0.119 loss_cls: 0.061 loss_box_reg: 0.058 time: 0.2304 data_time: 0.0115 lr: 0.000001 max_mem: 10044M [08/07 15:10:23 d2.utils.events]: eta: 0:29:34 iter: 7379 total_loss: 0.151 loss_cls: 0.073 loss_box_reg: 0.076 time: 0.2304 data_time: 0.0131 lr: 0.000001 max_mem: 10044M [08/07 15:10:28 d2.utils.events]: eta: 0:29:29 iter: 7399 total_loss: 0.135 loss_cls: 0.064 loss_box_reg: 0.065 time: 0.2304 data_time: 0.0109 lr: 0.000001 max_mem: 10044M [08/07 15:10:32 d2.utils.events]: eta: 0:29:25 iter: 7419 total_loss: 0.157 loss_cls: 0.081 loss_box_reg: 0.077 time: 0.2304 data_time: 0.0123 lr: 0.000001 max_mem: 10044M [08/07 15:10:37 d2.utils.events]: eta: 0:29:20 iter: 7439 total_loss: 0.138 loss_cls: 0.074 loss_box_reg: 0.064 time: 0.2304 data_time: 0.0148 lr: 0.000001 max_mem: 10044M [08/07 15:10:42 d2.utils.events]: eta: 0:29:15 iter: 7459 total_loss: 0.144 loss_cls: 0.074 loss_box_reg: 0.069 time: 0.2303 data_time: 0.0135 lr: 0.000001 max_mem: 10044M [08/07 15:10:46 d2.utils.events]: eta: 0:29:09 iter: 7479 total_loss: 0.148 loss_cls: 0.073 loss_box_reg: 0.073 time: 0.2303 data_time: 0.0120 lr: 0.000001 max_mem: 10044M [08/07 15:10:51 d2.utils.events]: eta: 0:29:02 iter: 7499 total_loss: 0.139 loss_cls: 0.073 loss_box_reg: 0.067 time: 0.2303 data_time: 0.0125 lr: 0.000001 max_mem: 10044M [08/07 15:10:55 d2.utils.events]: eta: 0:28:55 iter: 7519 total_loss: 0.155 loss_cls: 0.072 loss_box_reg: 0.076 time: 0.2303 data_time: 0.0128 lr: 0.000001 max_mem: 10044M [08/07 15:11:00 d2.utils.events]: eta: 0:28:48 iter: 7539 total_loss: 0.138 loss_cls: 0.074 loss_box_reg: 0.063 time: 0.2302 data_time: 0.0134 lr: 0.000001 max_mem: 10044M [08/07 15:11:04 d2.utils.events]: eta: 0:28:37 iter: 7559 total_loss: 0.146 loss_cls: 0.075 loss_box_reg: 0.070 time: 0.2302 data_time: 0.0138 lr: 0.000001 max_mem: 10044M [08/07 15:11:09 d2.utils.events]: eta: 0:28:32 iter: 7579 total_loss: 0.152 loss_cls: 0.076 loss_box_reg: 0.074 time: 0.2302 data_time: 0.0122 lr: 0.000001 max_mem: 10044M [08/07 15:11:13 d2.utils.events]: eta: 0:28:25 iter: 7599 total_loss: 0.136 loss_cls: 0.069 loss_box_reg: 0.066 time: 0.2301 data_time: 0.0151 lr: 0.000001 max_mem: 10044M [08/07 15:11:18 d2.utils.events]: eta: 0:28:19 iter: 7619 total_loss: 0.134 loss_cls: 0.071 loss_box_reg: 0.064 time: 0.2301 data_time: 0.0129 lr: 0.000001 max_mem: 10044M [08/07 15:11:23 d2.utils.events]: eta: 0:28:16 iter: 7639 total_loss: 0.129 loss_cls: 0.067 loss_box_reg: 0.063 time: 0.2301 data_time: 0.0154 lr: 0.000001 max_mem: 10044M [08/07 15:11:27 d2.utils.events]: eta: 0:28:12 iter: 7659 total_loss: 0.142 loss_cls: 0.074 loss_box_reg: 0.067 time: 0.2301 data_time: 0.0121 lr: 0.000001 max_mem: 10044M [08/07 15:11:32 d2.utils.events]: eta: 0:28:06 iter: 7679 total_loss: 0.134 loss_cls: 0.067 loss_box_reg: 0.064 time: 0.2300 data_time: 0.0112 lr: 0.000001 max_mem: 10044M [08/07 15:11:36 d2.utils.events]: eta: 0:28:03 iter: 7699 total_loss: 0.140 loss_cls: 0.073 loss_box_reg: 0.067 time: 0.2300 data_time: 0.0127 lr: 0.000001 max_mem: 10044M [08/07 15:11:41 d2.utils.events]: eta: 0:27:57 iter: 7719 total_loss: 0.120 loss_cls: 0.062 loss_box_reg: 0.055 time: 0.2300 data_time: 0.0145 lr: 0.000001 max_mem: 10044M [08/07 15:11:45 d2.utils.events]: eta: 0:27:47 iter: 7739 total_loss: 0.165 loss_cls: 0.082 loss_box_reg: 0.081 time: 0.2300 data_time: 0.0109 lr: 0.000001 max_mem: 10044M [08/07 15:11:50 d2.utils.events]: eta: 0:27:42 iter: 7759 total_loss: 0.151 loss_cls: 0.077 loss_box_reg: 0.073 time: 0.2299 data_time: 0.0155 lr: 0.000001 max_mem: 10044M [08/07 15:11:54 d2.utils.events]: eta: 0:27:34 iter: 7779 total_loss: 0.137 loss_cls: 0.072 loss_box_reg: 0.066 time: 0.2299 data_time: 0.0129 lr: 0.000001 max_mem: 10044M [08/07 15:11:59 d2.utils.events]: eta: 0:27:29 iter: 7799 total_loss: 0.117 loss_cls: 0.062 loss_box_reg: 0.056 time: 0.2299 data_time: 0.0137 lr: 0.000001 max_mem: 10044M [08/07 15:12:03 d2.utils.events]: eta: 0:27:24 iter: 7819 total_loss: 0.149 loss_cls: 0.075 loss_box_reg: 0.070 time: 0.2299 data_time: 0.0134 lr: 0.000001 max_mem: 10044M [08/07 15:12:08 d2.utils.events]: eta: 0:27:19 iter: 7839 total_loss: 0.139 loss_cls: 0.071 loss_box_reg: 0.067 time: 0.2298 data_time: 0.0147 lr: 0.000001 max_mem: 10044M [08/07 15:12:13 d2.utils.events]: eta: 0:27:14 iter: 7859 total_loss: 0.142 loss_cls: 0.075 loss_box_reg: 0.067 time: 0.2299 data_time: 0.0112 lr: 0.000001 max_mem: 10044M [08/07 15:12:17 d2.utils.events]: eta: 0:27:09 iter: 7879 total_loss: 0.140 loss_cls: 0.074 loss_box_reg: 0.065 time: 0.2298 data_time: 0.0127 lr: 0.000001 max_mem: 10044M [08/07 15:12:22 d2.utils.events]: eta: 0:27:04 iter: 7899 total_loss: 0.144 loss_cls: 0.072 loss_box_reg: 0.071 time: 0.2298 data_time: 0.0126 lr: 0.000001 max_mem: 10044M [08/07 15:12:26 d2.utils.events]: eta: 0:26:59 iter: 7919 total_loss: 0.143 loss_cls: 0.075 loss_box_reg: 0.071 time: 0.2298 data_time: 0.0123 lr: 0.000001 max_mem: 10044M [08/07 15:12:31 d2.utils.events]: eta: 0:26:55 iter: 7939 total_loss: 0.142 loss_cls: 0.071 loss_box_reg: 0.069 time: 0.2298 data_time: 0.0108 lr: 0.000001 max_mem: 10044M [08/07 15:12:35 d2.utils.events]: eta: 0:26:49 iter: 7959 total_loss: 0.136 loss_cls: 0.069 loss_box_reg: 0.066 time: 0.2297 data_time: 0.0122 lr: 0.000001 max_mem: 10044M [08/07 15:12:40 d2.utils.events]: eta: 0:26:42 iter: 7979 total_loss: 0.133 loss_cls: 0.072 loss_box_reg: 0.067 time: 0.2296 data_time: 0.0148 lr: 0.000001 max_mem: 10044M [08/07 15:12:44 d2.utils.events]: eta: 0:26:39 iter: 7999 total_loss: 0.146 loss_cls: 0.073 loss_box_reg: 0.072 time: 0.2296 data_time: 0.0121 lr: 0.000001 max_mem: 10044M [08/07 15:12:49 d2.utils.events]: eta: 0:26:34 iter: 8019 total_loss: 0.146 loss_cls: 0.075 loss_box_reg: 0.070 time: 0.2296 data_time: 0.0119 lr: 0.000001 max_mem: 10044M [08/07 15:12:53 d2.utils.events]: eta: 0:26:29 iter: 8039 total_loss: 0.122 loss_cls: 0.065 loss_box_reg: 0.057 time: 0.2296 data_time: 0.0127 lr: 0.000001 max_mem: 10044M [08/07 15:12:58 d2.utils.events]: eta: 0:26:24 iter: 8059 total_loss: 0.128 loss_cls: 0.068 loss_box_reg: 0.061 time: 0.2296 data_time: 0.0134 lr: 0.000001 max_mem: 10044M [08/07 15:13:03 d2.utils.events]: eta: 0:26:19 iter: 8079 total_loss: 0.141 loss_cls: 0.071 loss_box_reg: 0.067 time: 0.2296 data_time: 0.0106 lr: 0.000001 max_mem: 10044M [08/07 15:13:07 d2.utils.events]: eta: 0:26:15 iter: 8099 total_loss: 0.150 loss_cls: 0.077 loss_box_reg: 0.071 time: 0.2295 data_time: 0.0130 lr: 0.000001 max_mem: 10044M [08/07 15:13:12 d2.utils.events]: eta: 0:26:10 iter: 8119 total_loss: 0.133 loss_cls: 0.071 loss_box_reg: 0.061 time: 0.2295 data_time: 0.0127 lr: 0.000001 max_mem: 10044M [08/07 15:13:16 d2.utils.events]: eta: 0:26:06 iter: 8139 total_loss: 0.140 loss_cls: 0.073 loss_box_reg: 0.065 time: 0.2295 data_time: 0.0147 lr: 0.000001 max_mem: 10044M [08/07 15:13:21 d2.utils.events]: eta: 0:26:01 iter: 8159 total_loss: 0.142 loss_cls: 0.074 loss_box_reg: 0.068 time: 0.2295 data_time: 0.0129 lr: 0.000001 max_mem: 10044M [08/07 15:13:25 d2.utils.events]: eta: 0:25:56 iter: 8179 total_loss: 0.136 loss_cls: 0.070 loss_box_reg: 0.068 time: 0.2295 data_time: 0.0103 lr: 0.000001 max_mem: 10044M [08/07 15:13:30 d2.utils.events]: eta: 0:25:52 iter: 8199 total_loss: 0.143 loss_cls: 0.076 loss_box_reg: 0.066 time: 0.2295 data_time: 0.0103 lr: 0.000001 max_mem: 10044M [08/07 15:13:34 d2.utils.events]: eta: 0:25:46 iter: 8219 total_loss: 0.131 loss_cls: 0.070 loss_box_reg: 0.062 time: 0.2294 data_time: 0.0116 lr: 0.000001 max_mem: 10044M [08/07 15:13:39 d2.utils.events]: eta: 0:25:42 iter: 8239 total_loss: 0.135 loss_cls: 0.067 loss_box_reg: 0.066 time: 0.2294 data_time: 0.0121 lr: 0.000001 max_mem: 10044M [08/07 15:13:43 d2.utils.events]: eta: 0:25:37 iter: 8259 total_loss: 0.136 loss_cls: 0.068 loss_box_reg: 0.072 time: 0.2294 data_time: 0.0129 lr: 0.000001 max_mem: 10044M [08/07 15:13:48 d2.utils.events]: eta: 0:25:32 iter: 8279 total_loss: 0.144 loss_cls: 0.073 loss_box_reg: 0.068 time: 0.2294 data_time: 0.0113 lr: 0.000001 max_mem: 10044M [08/07 15:13:53 d2.utils.events]: eta: 0:25:27 iter: 8299 total_loss: 0.130 loss_cls: 0.070 loss_box_reg: 0.063 time: 0.2294 data_time: 0.0126 lr: 0.000001 max_mem: 10044M [08/07 15:13:57 d2.utils.events]: eta: 0:25:22 iter: 8319 total_loss: 0.140 loss_cls: 0.074 loss_box_reg: 0.068 time: 0.2293 data_time: 0.0142 lr: 0.000001 max_mem: 10044M [08/07 15:14:02 d2.utils.events]: eta: 0:25:15 iter: 8339 total_loss: 0.122 loss_cls: 0.063 loss_box_reg: 0.061 time: 0.2293 data_time: 0.0111 lr: 0.000001 max_mem: 10044M [08/07 15:14:06 d2.utils.events]: eta: 0:25:14 iter: 8359 total_loss: 0.148 loss_cls: 0.076 loss_box_reg: 0.072 time: 0.2293 data_time: 0.0174 lr: 0.000001 max_mem: 10044M [08/07 15:14:11 d2.utils.events]: eta: 0:25:10 iter: 8379 total_loss: 0.123 loss_cls: 0.067 loss_box_reg: 0.061 time: 0.2293 data_time: 0.0121 lr: 0.000001 max_mem: 10044M [08/07 15:14:16 d2.utils.events]: eta: 0:25:06 iter: 8399 total_loss: 0.145 loss_cls: 0.074 loss_box_reg: 0.069 time: 0.2294 data_time: 0.0140 lr: 0.000001 max_mem: 10044M [08/07 15:14:20 d2.utils.events]: eta: 0:25:02 iter: 8419 total_loss: 0.136 loss_cls: 0.068 loss_box_reg: 0.068 time: 0.2294 data_time: 0.0161 lr: 0.000001 max_mem: 10044M [08/07 15:14:25 d2.utils.events]: eta: 0:24:56 iter: 8439 total_loss: 0.127 loss_cls: 0.067 loss_box_reg: 0.060 time: 0.2294 data_time: 0.0195 lr: 0.000001 max_mem: 10044M [08/07 15:14:29 d2.utils.events]: eta: 0:24:53 iter: 8459 total_loss: 0.154 loss_cls: 0.078 loss_box_reg: 0.077 time: 0.2294 data_time: 0.0127 lr: 0.000001 max_mem: 10044M [08/07 15:14:34 d2.utils.events]: eta: 0:24:47 iter: 8479 total_loss: 0.132 loss_cls: 0.067 loss_box_reg: 0.065 time: 0.2294 data_time: 0.0113 lr: 0.000001 max_mem: 10044M [08/07 15:14:39 d2.utils.events]: eta: 0:24:42 iter: 8499 total_loss: 0.140 loss_cls: 0.075 loss_box_reg: 0.068 time: 0.2293 data_time: 0.0113 lr: 0.000001 max_mem: 10044M [08/07 15:14:43 d2.utils.events]: eta: 0:24:37 iter: 8519 total_loss: 0.145 loss_cls: 0.075 loss_box_reg: 0.070 time: 0.2293 data_time: 0.0114 lr: 0.000001 max_mem: 10044M [08/07 15:14:48 d2.utils.events]: eta: 0:24:34 iter: 8539 total_loss: 0.141 loss_cls: 0.067 loss_box_reg: 0.071 time: 0.2293 data_time: 0.0117 lr: 0.000001 max_mem: 10044M [08/07 15:14:52 d2.utils.events]: eta: 0:24:30 iter: 8559 total_loss: 0.126 loss_cls: 0.067 loss_box_reg: 0.062 time: 0.2293 data_time: 0.0126 lr: 0.000001 max_mem: 10044M [08/07 15:14:57 d2.utils.events]: eta: 0:24:25 iter: 8579 total_loss: 0.154 loss_cls: 0.082 loss_box_reg: 0.072 time: 0.2294 data_time: 0.0133 lr: 0.000001 max_mem: 10044M [08/07 15:15:01 d2.utils.events]: eta: 0:24:21 iter: 8599 total_loss: 0.142 loss_cls: 0.072 loss_box_reg: 0.066 time: 0.2293 data_time: 0.0133 lr: 0.000001 max_mem: 10044M [08/07 15:15:06 d2.utils.events]: eta: 0:24:17 iter: 8619 total_loss: 0.139 loss_cls: 0.071 loss_box_reg: 0.069 time: 0.2293 data_time: 0.0146 lr: 0.000001 max_mem: 10044M [08/07 15:15:11 d2.utils.events]: eta: 0:24:12 iter: 8639 total_loss: 0.132 loss_cls: 0.068 loss_box_reg: 0.069 time: 0.2293 data_time: 0.0116 lr: 0.000001 max_mem: 10044M [08/07 15:15:15 d2.utils.events]: eta: 0:24:09 iter: 8659 total_loss: 0.147 loss_cls: 0.074 loss_box_reg: 0.073 time: 0.2293 data_time: 0.0125 lr: 0.000001 max_mem: 10044M [08/07 15:15:20 d2.utils.events]: eta: 0:24:04 iter: 8679 total_loss: 0.139 loss_cls: 0.072 loss_box_reg: 0.067 time: 0.2293 data_time: 0.0118 lr: 0.000001 max_mem: 10044M [08/07 15:15:25 d2.utils.events]: eta: 0:23:59 iter: 8699 total_loss: 0.132 loss_cls: 0.069 loss_box_reg: 0.060 time: 0.2294 data_time: 0.0150 lr: 0.000001 max_mem: 10044M [08/07 15:15:30 d2.utils.events]: eta: 0:23:56 iter: 8719 total_loss: 0.133 loss_cls: 0.065 loss_box_reg: 0.068 time: 0.2295 data_time: 0.0139 lr: 0.000001 max_mem: 10044M [08/07 15:15:34 d2.utils.events]: eta: 0:23:54 iter: 8739 total_loss: 0.143 loss_cls: 0.074 loss_box_reg: 0.069 time: 0.2294 data_time: 0.0129 lr: 0.000001 max_mem: 10044M [08/07 15:15:39 d2.utils.events]: eta: 0:23:50 iter: 8759 total_loss: 0.141 loss_cls: 0.073 loss_box_reg: 0.070 time: 0.2294 data_time: 0.0121 lr: 0.000001 max_mem: 10044M [08/07 15:15:43 d2.utils.events]: eta: 0:23:45 iter: 8779 total_loss: 0.134 loss_cls: 0.069 loss_box_reg: 0.067 time: 0.2294 data_time: 0.0112 lr: 0.000001 max_mem: 10044M [08/07 15:15:48 d2.utils.events]: eta: 0:23:42 iter: 8799 total_loss: 0.144 loss_cls: 0.074 loss_box_reg: 0.068 time: 0.2294 data_time: 0.0118 lr: 0.000001 max_mem: 10044M [08/07 15:15:53 d2.utils.events]: eta: 0:23:37 iter: 8819 total_loss: 0.132 loss_cls: 0.068 loss_box_reg: 0.064 time: 0.2294 data_time: 0.0115 lr: 0.000001 max_mem: 10044M [08/07 15:15:57 d2.utils.events]: eta: 0:23:34 iter: 8839 total_loss: 0.147 loss_cls: 0.073 loss_box_reg: 0.070 time: 0.2294 data_time: 0.0149 lr: 0.000001 max_mem: 10044M [08/07 15:16:02 d2.utils.events]: eta: 0:23:28 iter: 8859 total_loss: 0.138 loss_cls: 0.073 loss_box_reg: 0.067 time: 0.2294 data_time: 0.0149 lr: 0.000001 max_mem: 10044M [08/07 15:16:06 d2.utils.events]: eta: 0:23:24 iter: 8879 total_loss: 0.127 loss_cls: 0.066 loss_box_reg: 0.059 time: 0.2294 data_time: 0.0110 lr: 0.000001 max_mem: 10044M [08/07 15:16:11 d2.utils.events]: eta: 0:23:19 iter: 8899 total_loss: 0.144 loss_cls: 0.074 loss_box_reg: 0.072 time: 0.2294 data_time: 0.0130 lr: 0.000001 max_mem: 10044M [08/07 15:16:15 d2.utils.events]: eta: 0:23:13 iter: 8919 total_loss: 0.140 loss_cls: 0.072 loss_box_reg: 0.069 time: 0.2293 data_time: 0.0105 lr: 0.000001 max_mem: 10044M [08/07 15:16:20 d2.utils.events]: eta: 0:23:05 iter: 8939 total_loss: 0.139 loss_cls: 0.074 loss_box_reg: 0.065 time: 0.2293 data_time: 0.0123 lr: 0.000001 max_mem: 10044M [08/07 15:16:24 d2.utils.events]: eta: 0:23:04 iter: 8959 total_loss: 0.144 loss_cls: 0.074 loss_box_reg: 0.071 time: 0.2293 data_time: 0.0141 lr: 0.000001 max_mem: 10044M [08/07 15:16:29 d2.utils.events]: eta: 0:22:58 iter: 8979 total_loss: 0.140 loss_cls: 0.068 loss_box_reg: 0.071 time: 0.2293 data_time: 0.0149 lr: 0.000001 max_mem: 10044M [08/07 15:16:33 d2.utils.events]: eta: 0:22:54 iter: 8999 total_loss: 0.134 loss_cls: 0.072 loss_box_reg: 0.060 time: 0.2293 data_time: 0.0118 lr: 0.000001 max_mem: 10044M [08/07 15:16:38 d2.utils.events]: eta: 0:22:50 iter: 9019 total_loss: 0.137 loss_cls: 0.069 loss_box_reg: 0.068 time: 0.2293 data_time: 0.0150 lr: 0.000001 max_mem: 10044M [08/07 15:16:42 d2.utils.events]: eta: 0:22:43 iter: 9039 total_loss: 0.139 loss_cls: 0.073 loss_box_reg: 0.067 time: 0.2293 data_time: 0.0107 lr: 0.000001 max_mem: 10044M [08/07 15:16:47 d2.utils.events]: eta: 0:22:40 iter: 9059 total_loss: 0.142 loss_cls: 0.074 loss_box_reg: 0.067 time: 0.2292 data_time: 0.0147 lr: 0.000001 max_mem: 10044M [08/07 15:16:52 d2.utils.events]: eta: 0:22:37 iter: 9079 total_loss: 0.141 loss_cls: 0.074 loss_box_reg: 0.068 time: 0.2292 data_time: 0.0121 lr: 0.000001 max_mem: 10044M [08/07 15:16:56 d2.utils.events]: eta: 0:22:32 iter: 9099 total_loss: 0.131 loss_cls: 0.071 loss_box_reg: 0.060 time: 0.2293 data_time: 0.0108 lr: 0.000001 max_mem: 10044M [08/07 15:17:01 d2.utils.events]: eta: 0:22:26 iter: 9119 total_loss: 0.127 loss_cls: 0.065 loss_box_reg: 0.061 time: 0.2293 data_time: 0.0110 lr: 0.000001 max_mem: 10044M [08/07 15:17:05 d2.utils.events]: eta: 0:22:20 iter: 9139 total_loss: 0.140 loss_cls: 0.075 loss_box_reg: 0.066 time: 0.2292 data_time: 0.0119 lr: 0.000001 max_mem: 10044M [08/07 15:17:10 d2.utils.events]: eta: 0:22:14 iter: 9159 total_loss: 0.129 loss_cls: 0.064 loss_box_reg: 0.059 time: 0.2292 data_time: 0.0122 lr: 0.000001 max_mem: 10044M [08/07 15:17:14 d2.utils.events]: eta: 0:22:12 iter: 9179 total_loss: 0.138 loss_cls: 0.072 loss_box_reg: 0.067 time: 0.2292 data_time: 0.0118 lr: 0.000001 max_mem: 10044M [08/07 15:17:19 d2.utils.events]: eta: 0:22:09 iter: 9199 total_loss: 0.139 loss_cls: 0.072 loss_box_reg: 0.066 time: 0.2292 data_time: 0.0146 lr: 0.000001 max_mem: 10044M [08/07 15:17:24 d2.utils.events]: eta: 0:22:03 iter: 9219 total_loss: 0.143 loss_cls: 0.071 loss_box_reg: 0.070 time: 0.2292 data_time: 0.0122 lr: 0.000001 max_mem: 10044M [08/07 15:17:28 d2.utils.events]: eta: 0:21:57 iter: 9239 total_loss: 0.141 loss_cls: 0.075 loss_box_reg: 0.066 time: 0.2291 data_time: 0.0113 lr: 0.000001 max_mem: 10044M [08/07 15:17:32 d2.utils.events]: eta: 0:21:52 iter: 9259 total_loss: 0.145 loss_cls: 0.074 loss_box_reg: 0.069 time: 0.2291 data_time: 0.0108 lr: 0.000001 max_mem: 10044M [08/07 15:17:37 d2.utils.events]: eta: 0:21:46 iter: 9279 total_loss: 0.138 loss_cls: 0.073 loss_box_reg: 0.066 time: 0.2291 data_time: 0.0103 lr: 0.000001 max_mem: 10044M [08/07 15:17:42 d2.utils.events]: eta: 0:21:42 iter: 9299 total_loss: 0.160 loss_cls: 0.079 loss_box_reg: 0.077 time: 0.2291 data_time: 0.0134 lr: 0.000001 max_mem: 10044M [08/07 15:17:46 d2.utils.events]: eta: 0:21:36 iter: 9319 total_loss: 0.133 loss_cls: 0.068 loss_box_reg: 0.064 time: 0.2291 data_time: 0.0129 lr: 0.000001 max_mem: 10044M [08/07 15:17:50 d2.utils.events]: eta: 0:21:32 iter: 9339 total_loss: 0.143 loss_cls: 0.071 loss_box_reg: 0.069 time: 0.2290 data_time: 0.0119 lr: 0.000001 max_mem: 10044M [08/07 15:17:55 d2.utils.events]: eta: 0:21:29 iter: 9359 total_loss: 0.131 loss_cls: 0.066 loss_box_reg: 0.066 time: 0.2290 data_time: 0.0149 lr: 0.000001 max_mem: 10044M [08/07 15:18:00 d2.utils.events]: eta: 0:21:23 iter: 9379 total_loss: 0.140 loss_cls: 0.072 loss_box_reg: 0.066 time: 0.2290 data_time: 0.0117 lr: 0.000001 max_mem: 10044M [08/07 15:18:04 d2.utils.events]: eta: 0:21:18 iter: 9399 total_loss: 0.141 loss_cls: 0.075 loss_box_reg: 0.069 time: 0.2290 data_time: 0.0151 lr: 0.000001 max_mem: 10044M [08/07 15:18:09 d2.utils.events]: eta: 0:21:10 iter: 9419 total_loss: 0.136 loss_cls: 0.068 loss_box_reg: 0.066 time: 0.2290 data_time: 0.0122 lr: 0.000001 max_mem: 10044M [08/07 15:18:13 d2.utils.events]: eta: 0:21:05 iter: 9439 total_loss: 0.130 loss_cls: 0.067 loss_box_reg: 0.061 time: 0.2290 data_time: 0.0117 lr: 0.000001 max_mem: 10044M [08/07 15:18:18 d2.utils.events]: eta: 0:20:58 iter: 9459 total_loss: 0.129 loss_cls: 0.066 loss_box_reg: 0.065 time: 0.2290 data_time: 0.0114 lr: 0.000001 max_mem: 10044M [08/07 15:18:22 d2.utils.events]: eta: 0:20:56 iter: 9479 total_loss: 0.144 loss_cls: 0.074 loss_box_reg: 0.072 time: 0.2289 data_time: 0.0127 lr: 0.000001 max_mem: 10044M [08/07 15:18:27 d2.utils.events]: eta: 0:20:49 iter: 9499 total_loss: 0.131 loss_cls: 0.067 loss_box_reg: 0.063 time: 0.2289 data_time: 0.0108 lr: 0.000001 max_mem: 10044M [08/07 15:18:31 d2.utils.events]: eta: 0:20:46 iter: 9519 total_loss: 0.149 loss_cls: 0.072 loss_box_reg: 0.074 time: 0.2289 data_time: 0.0110 lr: 0.000001 max_mem: 10044M [08/07 15:18:36 d2.utils.events]: eta: 0:20:39 iter: 9539 total_loss: 0.141 loss_cls: 0.072 loss_box_reg: 0.066 time: 0.2289 data_time: 0.0138 lr: 0.000001 max_mem: 10044M [08/07 15:18:40 d2.utils.events]: eta: 0:20:33 iter: 9559 total_loss: 0.123 loss_cls: 0.064 loss_box_reg: 0.061 time: 0.2289 data_time: 0.0128 lr: 0.000001 max_mem: 10044M [08/07 15:18:45 d2.utils.events]: eta: 0:20:29 iter: 9579 total_loss: 0.125 loss_cls: 0.065 loss_box_reg: 0.062 time: 0.2289 data_time: 0.0141 lr: 0.000001 max_mem: 10044M [08/07 15:18:50 d2.utils.events]: eta: 0:20:25 iter: 9599 total_loss: 0.143 loss_cls: 0.073 loss_box_reg: 0.067 time: 0.2289 data_time: 0.0117 lr: 0.000001 max_mem: 10044M [08/07 15:18:54 d2.utils.events]: eta: 0:20:20 iter: 9619 total_loss: 0.131 loss_cls: 0.071 loss_box_reg: 0.066 time: 0.2289 data_time: 0.0117 lr: 0.000001 max_mem: 10044M [08/07 15:18:59 d2.utils.events]: eta: 0:20:16 iter: 9639 total_loss: 0.129 loss_cls: 0.068 loss_box_reg: 0.060 time: 0.2289 data_time: 0.0119 lr: 0.000001 max_mem: 10044M [08/07 15:19:03 d2.utils.events]: eta: 0:20:10 iter: 9659 total_loss: 0.134 loss_cls: 0.069 loss_box_reg: 0.063 time: 0.2289 data_time: 0.0131 lr: 0.000001 max_mem: 10044M [08/07 15:19:08 d2.utils.events]: eta: 0:20:05 iter: 9679 total_loss: 0.142 loss_cls: 0.074 loss_box_reg: 0.072 time: 0.2288 data_time: 0.0115 lr: 0.000001 max_mem: 10044M [08/07 15:19:12 d2.utils.events]: eta: 0:20:00 iter: 9699 total_loss: 0.132 loss_cls: 0.068 loss_box_reg: 0.064 time: 0.2288 data_time: 0.0109 lr: 0.000001 max_mem: 10044M [08/07 15:19:17 d2.utils.events]: eta: 0:19:53 iter: 9719 total_loss: 0.135 loss_cls: 0.071 loss_box_reg: 0.066 time: 0.2288 data_time: 0.0112 lr: 0.000001 max_mem: 10044M [08/07 15:19:21 d2.utils.events]: eta: 0:19:49 iter: 9739 total_loss: 0.138 loss_cls: 0.073 loss_box_reg: 0.065 time: 0.2288 data_time: 0.0122 lr: 0.000001 max_mem: 10044M [08/07 15:19:26 d2.utils.events]: eta: 0:19:45 iter: 9759 total_loss: 0.147 loss_cls: 0.080 loss_box_reg: 0.067 time: 0.2288 data_time: 0.0127 lr: 0.000001 max_mem: 10044M [08/07 15:19:31 d2.utils.events]: eta: 0:19:40 iter: 9779 total_loss: 0.153 loss_cls: 0.079 loss_box_reg: 0.074 time: 0.2288 data_time: 0.0139 lr: 0.000001 max_mem: 10044M [08/07 15:19:35 d2.utils.events]: eta: 0:19:35 iter: 9799 total_loss: 0.139 loss_cls: 0.068 loss_box_reg: 0.070 time: 0.2288 data_time: 0.0134 lr: 0.000001 max_mem: 10044M [08/07 15:19:40 d2.utils.events]: eta: 0:19:31 iter: 9819 total_loss: 0.141 loss_cls: 0.074 loss_box_reg: 0.070 time: 0.2288 data_time: 0.0146 lr: 0.000001 max_mem: 10044M [08/07 15:19:44 d2.utils.events]: eta: 0:19:26 iter: 9839 total_loss: 0.149 loss_cls: 0.076 loss_box_reg: 0.072 time: 0.2288 data_time: 0.0108 lr: 0.000001 max_mem: 10044M [08/07 15:19:49 d2.utils.events]: eta: 0:19:23 iter: 9859 total_loss: 0.136 loss_cls: 0.072 loss_box_reg: 0.067 time: 0.2288 data_time: 0.0118 lr: 0.000001 max_mem: 10044M [08/07 15:19:54 d2.utils.events]: eta: 0:19:19 iter: 9879 total_loss: 0.135 loss_cls: 0.073 loss_box_reg: 0.067 time: 0.2288 data_time: 0.0127 lr: 0.000001 max_mem: 10044M [08/07 15:19:58 d2.utils.events]: eta: 0:19:15 iter: 9899 total_loss: 0.148 loss_cls: 0.075 loss_box_reg: 0.075 time: 0.2288 data_time: 0.0140 lr: 0.000001 max_mem: 10044M [08/07 15:20:03 d2.utils.events]: eta: 0:19:12 iter: 9919 total_loss: 0.133 loss_cls: 0.066 loss_box_reg: 0.065 time: 0.2288 data_time: 0.0123 lr: 0.000001 max_mem: 10044M [08/07 15:20:07 d2.utils.events]: eta: 0:19:08 iter: 9939 total_loss: 0.145 loss_cls: 0.077 loss_box_reg: 0.069 time: 0.2288 data_time: 0.0105 lr: 0.000001 max_mem: 10044M [08/07 15:20:12 d2.utils.events]: eta: 0:19:04 iter: 9959 total_loss: 0.152 loss_cls: 0.074 loss_box_reg: 0.075 time: 0.2288 data_time: 0.0157 lr: 0.000001 max_mem: 10044M [08/07 15:20:16 d2.utils.events]: eta: 0:19:00 iter: 9979 total_loss: 0.145 loss_cls: 0.075 loss_box_reg: 0.069 time: 0.2288 data_time: 0.0134 lr: 0.000001 max_mem: 10044M [08/07 15:20:23 d2.utils.events]: eta: 0:18:56 iter: 9999 total_loss: 0.147 loss_cls: 0.074 loss_box_reg: 0.069 time: 0.2288 data_time: 0.0118 lr: 0.000001 max_mem: 10044M [08/07 15:20:28 d2.utils.events]: eta: 0:18:50 iter: 10019 total_loss: 0.150 loss_cls: 0.080 loss_box_reg: 0.073 time: 0.2287 data_time: 0.0117 lr: 0.000000 max_mem: 10044M [08/07 15:20:33 d2.utils.events]: eta: 0:18:45 iter: 10039 total_loss: 0.132 loss_cls: 0.073 loss_box_reg: 0.066 time: 0.2288 data_time: 0.0132 lr: 0.000000 max_mem: 10044M [08/07 15:20:38 d2.utils.events]: eta: 0:18:42 iter: 10059 total_loss: 0.132 loss_cls: 0.070 loss_box_reg: 0.064 time: 0.2289 data_time: 0.0163 lr: 0.000000 max_mem: 10044M [08/07 15:20:42 d2.utils.events]: eta: 0:18:37 iter: 10079 total_loss: 0.143 loss_cls: 0.075 loss_box_reg: 0.070 time: 0.2289 data_time: 0.0149 lr: 0.000000 max_mem: 10044M [08/07 15:20:47 d2.utils.events]: eta: 0:18:32 iter: 10099 total_loss: 0.147 loss_cls: 0.074 loss_box_reg: 0.072 time: 0.2289 data_time: 0.0135 lr: 0.000000 max_mem: 10044M [08/07 15:20:52 d2.utils.events]: eta: 0:18:29 iter: 10119 total_loss: 0.138 loss_cls: 0.071 loss_box_reg: 0.067 time: 0.2289 data_time: 0.0141 lr: 0.000000 max_mem: 10044M [08/07 15:20:56 d2.utils.events]: eta: 0:18:26 iter: 10139 total_loss: 0.134 loss_cls: 0.070 loss_box_reg: 0.065 time: 0.2289 data_time: 0.0125 lr: 0.000000 max_mem: 10044M [08/07 15:21:01 d2.utils.events]: eta: 0:18:22 iter: 10159 total_loss: 0.143 loss_cls: 0.076 loss_box_reg: 0.066 time: 0.2289 data_time: 0.0129 lr: 0.000000 max_mem: 10044M [08/07 15:21:05 d2.utils.events]: eta: 0:18:17 iter: 10179 total_loss: 0.141 loss_cls: 0.074 loss_box_reg: 0.068 time: 0.2289 data_time: 0.0120 lr: 0.000000 max_mem: 10044M [08/07 15:21:10 d2.utils.events]: eta: 0:18:12 iter: 10199 total_loss: 0.137 loss_cls: 0.070 loss_box_reg: 0.064 time: 0.2289 data_time: 0.0124 lr: 0.000000 max_mem: 10044M [08/07 15:21:14 d2.utils.events]: eta: 0:18:08 iter: 10219 total_loss: 0.134 loss_cls: 0.069 loss_box_reg: 0.065 time: 0.2289 data_time: 0.0144 lr: 0.000000 max_mem: 10044M [08/07 15:21:19 d2.utils.events]: eta: 0:18:04 iter: 10239 total_loss: 0.145 loss_cls: 0.073 loss_box_reg: 0.069 time: 0.2289 data_time: 0.0130 lr: 0.000000 max_mem: 10044M [08/07 15:21:24 d2.utils.events]: eta: 0:18:02 iter: 10259 total_loss: 0.109 loss_cls: 0.061 loss_box_reg: 0.052 time: 0.2289 data_time: 0.0133 lr: 0.000000 max_mem: 10044M [08/07 15:21:28 d2.utils.events]: eta: 0:17:58 iter: 10279 total_loss: 0.124 loss_cls: 0.067 loss_box_reg: 0.060 time: 0.2289 data_time: 0.0126 lr: 0.000000 max_mem: 10044M [08/07 15:21:33 d2.utils.events]: eta: 0:17:53 iter: 10299 total_loss: 0.134 loss_cls: 0.070 loss_box_reg: 0.062 time: 0.2288 data_time: 0.0093 lr: 0.000000 max_mem: 10044M [08/07 15:21:37 d2.utils.events]: eta: 0:17:46 iter: 10319 total_loss: 0.131 loss_cls: 0.071 loss_box_reg: 0.065 time: 0.2288 data_time: 0.0109 lr: 0.000000 max_mem: 10044M [08/07 15:21:42 d2.utils.events]: eta: 0:17:43 iter: 10339 total_loss: 0.143 loss_cls: 0.077 loss_box_reg: 0.065 time: 0.2288 data_time: 0.0134 lr: 0.000000 max_mem: 10044M [08/07 15:21:46 d2.utils.events]: eta: 0:17:36 iter: 10359 total_loss: 0.136 loss_cls: 0.071 loss_box_reg: 0.062 time: 0.2288 data_time: 0.0113 lr: 0.000000 max_mem: 10044M [08/07 15:21:51 d2.utils.events]: eta: 0:17:31 iter: 10379 total_loss: 0.147 loss_cls: 0.078 loss_box_reg: 0.067 time: 0.2288 data_time: 0.0124 lr: 0.000000 max_mem: 10044M [08/07 15:21:55 d2.utils.events]: eta: 0:17:27 iter: 10399 total_loss: 0.147 loss_cls: 0.076 loss_box_reg: 0.069 time: 0.2288 data_time: 0.0140 lr: 0.000000 max_mem: 10044M [08/07 15:22:00 d2.utils.events]: eta: 0:17:23 iter: 10419 total_loss: 0.134 loss_cls: 0.069 loss_box_reg: 0.068 time: 0.2288 data_time: 0.0129 lr: 0.000000 max_mem: 10044M [08/07 15:22:04 d2.utils.events]: eta: 0:17:19 iter: 10439 total_loss: 0.143 loss_cls: 0.073 loss_box_reg: 0.070 time: 0.2287 data_time: 0.0125 lr: 0.000000 max_mem: 10044M [08/07 15:22:09 d2.utils.events]: eta: 0:17:17 iter: 10459 total_loss: 0.142 loss_cls: 0.073 loss_box_reg: 0.073 time: 0.2288 data_time: 0.0142 lr: 0.000000 max_mem: 10044M [08/07 15:22:14 d2.utils.events]: eta: 0:17:13 iter: 10479 total_loss: 0.146 loss_cls: 0.075 loss_box_reg: 0.072 time: 0.2288 data_time: 0.0166 lr: 0.000000 max_mem: 10044M [08/07 15:22:18 d2.utils.events]: eta: 0:17:09 iter: 10499 total_loss: 0.124 loss_cls: 0.064 loss_box_reg: 0.058 time: 0.2288 data_time: 0.0130 lr: 0.000000 max_mem: 10044M [08/07 15:22:23 d2.utils.events]: eta: 0:17:05 iter: 10519 total_loss: 0.129 loss_cls: 0.069 loss_box_reg: 0.062 time: 0.2288 data_time: 0.0129 lr: 0.000000 max_mem: 10044M [08/07 15:22:28 d2.utils.events]: eta: 0:17:00 iter: 10539 total_loss: 0.144 loss_cls: 0.071 loss_box_reg: 0.068 time: 0.2288 data_time: 0.0125 lr: 0.000000 max_mem: 10044M [08/07 15:22:32 d2.utils.events]: eta: 0:16:57 iter: 10559 total_loss: 0.132 loss_cls: 0.065 loss_box_reg: 0.067 time: 0.2288 data_time: 0.0104 lr: 0.000000 max_mem: 10044M [08/07 15:22:37 d2.utils.events]: eta: 0:16:51 iter: 10579 total_loss: 0.136 loss_cls: 0.071 loss_box_reg: 0.066 time: 0.2288 data_time: 0.0132 lr: 0.000000 max_mem: 10044M [08/07 15:22:41 d2.utils.events]: eta: 0:16:45 iter: 10599 total_loss: 0.133 loss_cls: 0.070 loss_box_reg: 0.063 time: 0.2287 data_time: 0.0107 lr: 0.000000 max_mem: 10044M [08/07 15:22:46 d2.utils.events]: eta: 0:16:40 iter: 10619 total_loss: 0.140 loss_cls: 0.072 loss_box_reg: 0.067 time: 0.2287 data_time: 0.0114 lr: 0.000000 max_mem: 10044M [08/07 15:22:50 d2.utils.events]: eta: 0:16:35 iter: 10639 total_loss: 0.146 loss_cls: 0.073 loss_box_reg: 0.069 time: 0.2287 data_time: 0.0116 lr: 0.000000 max_mem: 10044M [08/07 15:22:55 d2.utils.events]: eta: 0:16:31 iter: 10659 total_loss: 0.138 loss_cls: 0.071 loss_box_reg: 0.064 time: 0.2287 data_time: 0.0133 lr: 0.000000 max_mem: 10044M [08/07 15:22:59 d2.utils.events]: eta: 0:16:27 iter: 10679 total_loss: 0.127 loss_cls: 0.067 loss_box_reg: 0.061 time: 0.2287 data_time: 0.0126 lr: 0.000000 max_mem: 10044M [08/07 15:23:04 d2.utils.events]: eta: 0:16:22 iter: 10699 total_loss: 0.127 loss_cls: 0.066 loss_box_reg: 0.061 time: 0.2287 data_time: 0.0140 lr: 0.000000 max_mem: 10044M [08/07 15:23:08 d2.utils.events]: eta: 0:16:19 iter: 10719 total_loss: 0.145 loss_cls: 0.073 loss_box_reg: 0.070 time: 0.2287 data_time: 0.0127 lr: 0.000000 max_mem: 10044M [08/07 15:23:13 d2.utils.events]: eta: 0:16:14 iter: 10739 total_loss: 0.137 loss_cls: 0.070 loss_box_reg: 0.067 time: 0.2287 data_time: 0.0115 lr: 0.000000 max_mem: 10044M [08/07 15:23:17 d2.utils.events]: eta: 0:16:10 iter: 10759 total_loss: 0.126 loss_cls: 0.063 loss_box_reg: 0.066 time: 0.2287 data_time: 0.0126 lr: 0.000000 max_mem: 10044M [08/07 15:23:22 d2.utils.events]: eta: 0:16:05 iter: 10779 total_loss: 0.147 loss_cls: 0.074 loss_box_reg: 0.070 time: 0.2287 data_time: 0.0132 lr: 0.000000 max_mem: 10044M [08/07 15:23:26 d2.utils.events]: eta: 0:16:01 iter: 10799 total_loss: 0.145 loss_cls: 0.074 loss_box_reg: 0.070 time: 0.2286 data_time: 0.0129 lr: 0.000000 max_mem: 10044M [08/07 15:23:31 d2.utils.events]: eta: 0:15:56 iter: 10819 total_loss: 0.156 loss_cls: 0.078 loss_box_reg: 0.075 time: 0.2286 data_time: 0.0142 lr: 0.000000 max_mem: 10044M [08/07 15:23:35 d2.utils.events]: eta: 0:15:52 iter: 10839 total_loss: 0.156 loss_cls: 0.079 loss_box_reg: 0.075 time: 0.2286 data_time: 0.0114 lr: 0.000000 max_mem: 10044M [08/07 15:23:40 d2.utils.events]: eta: 0:15:47 iter: 10859 total_loss: 0.141 loss_cls: 0.072 loss_box_reg: 0.067 time: 0.2286 data_time: 0.0128 lr: 0.000000 max_mem: 10044M [08/07 15:23:44 d2.utils.events]: eta: 0:15:41 iter: 10879 total_loss: 0.138 loss_cls: 0.073 loss_box_reg: 0.071 time: 0.2286 data_time: 0.0111 lr: 0.000000 max_mem: 10044M [08/07 15:23:49 d2.utils.events]: eta: 0:15:37 iter: 10899 total_loss: 0.141 loss_cls: 0.071 loss_box_reg: 0.068 time: 0.2285 data_time: 0.0124 lr: 0.000000 max_mem: 10044M [08/07 15:23:53 d2.utils.events]: eta: 0:15:33 iter: 10919 total_loss: 0.150 loss_cls: 0.074 loss_box_reg: 0.072 time: 0.2285 data_time: 0.0132 lr: 0.000000 max_mem: 10044M [08/07 15:23:58 d2.utils.events]: eta: 0:15:28 iter: 10939 total_loss: 0.147 loss_cls: 0.077 loss_box_reg: 0.070 time: 0.2286 data_time: 0.0137 lr: 0.000000 max_mem: 10044M [08/07 15:24:03 d2.utils.events]: eta: 0:15:23 iter: 10959 total_loss: 0.125 loss_cls: 0.065 loss_box_reg: 0.061 time: 0.2286 data_time: 0.0121 lr: 0.000000 max_mem: 10044M [08/07 15:24:08 d2.utils.events]: eta: 0:15:20 iter: 10979 total_loss: 0.156 loss_cls: 0.080 loss_box_reg: 0.078 time: 0.2286 data_time: 0.0141 lr: 0.000000 max_mem: 10044M [08/07 15:24:12 d2.utils.events]: eta: 0:15:14 iter: 10999 total_loss: 0.149 loss_cls: 0.074 loss_box_reg: 0.071 time: 0.2286 data_time: 0.0137 lr: 0.000000 max_mem: 10044M [08/07 15:24:17 d2.utils.events]: eta: 0:15:10 iter: 11019 total_loss: 0.143 loss_cls: 0.071 loss_box_reg: 0.065 time: 0.2286 data_time: 0.0135 lr: 0.000000 max_mem: 10044M [08/07 15:24:21 d2.utils.events]: eta: 0:15:06 iter: 11039 total_loss: 0.140 loss_cls: 0.074 loss_box_reg: 0.063 time: 0.2286 data_time: 0.0130 lr: 0.000000 max_mem: 10044M [08/07 15:24:26 d2.utils.events]: eta: 0:15:00 iter: 11059 total_loss: 0.128 loss_cls: 0.068 loss_box_reg: 0.063 time: 0.2287 data_time: 0.0130 lr: 0.000000 max_mem: 10044M [08/07 15:24:31 d2.utils.events]: eta: 0:14:56 iter: 11079 total_loss: 0.129 loss_cls: 0.069 loss_box_reg: 0.063 time: 0.2287 data_time: 0.0145 lr: 0.000000 max_mem: 10044M [08/07 15:24:36 d2.utils.events]: eta: 0:14:52 iter: 11099 total_loss: 0.126 loss_cls: 0.066 loss_box_reg: 0.061 time: 0.2287 data_time: 0.0137 lr: 0.000000 max_mem: 10044M [08/07 15:24:40 d2.utils.events]: eta: 0:14:47 iter: 11119 total_loss: 0.138 loss_cls: 0.072 loss_box_reg: 0.067 time: 0.2287 data_time: 0.0151 lr: 0.000000 max_mem: 10044M [08/07 15:24:45 d2.utils.events]: eta: 0:14:42 iter: 11139 total_loss: 0.145 loss_cls: 0.076 loss_box_reg: 0.067 time: 0.2287 data_time: 0.0122 lr: 0.000000 max_mem: 10044M [08/07 15:24:50 d2.utils.events]: eta: 0:14:37 iter: 11159 total_loss: 0.152 loss_cls: 0.076 loss_box_reg: 0.075 time: 0.2287 data_time: 0.0131 lr: 0.000000 max_mem: 10044M [08/07 15:24:54 d2.utils.events]: eta: 0:14:34 iter: 11179 total_loss: 0.140 loss_cls: 0.066 loss_box_reg: 0.069 time: 0.2288 data_time: 0.0141 lr: 0.000000 max_mem: 10044M [08/07 15:24:59 d2.utils.events]: eta: 0:14:29 iter: 11199 total_loss: 0.144 loss_cls: 0.070 loss_box_reg: 0.071 time: 0.2288 data_time: 0.0132 lr: 0.000000 max_mem: 10044M [08/07 15:25:04 d2.utils.events]: eta: 0:14:25 iter: 11219 total_loss: 0.141 loss_cls: 0.068 loss_box_reg: 0.070 time: 0.2288 data_time: 0.0139 lr: 0.000000 max_mem: 10044M [08/07 15:25:08 d2.utils.events]: eta: 0:14:20 iter: 11239 total_loss: 0.128 loss_cls: 0.065 loss_box_reg: 0.059 time: 0.2287 data_time: 0.0128 lr: 0.000000 max_mem: 10044M [08/07 15:25:13 d2.utils.events]: eta: 0:14:15 iter: 11259 total_loss: 0.151 loss_cls: 0.076 loss_box_reg: 0.075 time: 0.2287 data_time: 0.0127 lr: 0.000000 max_mem: 10044M [08/07 15:25:17 d2.utils.events]: eta: 0:14:11 iter: 11279 total_loss: 0.134 loss_cls: 0.069 loss_box_reg: 0.064 time: 0.2287 data_time: 0.0124 lr: 0.000000 max_mem: 10044M [08/07 15:25:22 d2.utils.events]: eta: 0:14:06 iter: 11299 total_loss: 0.133 loss_cls: 0.067 loss_box_reg: 0.063 time: 0.2287 data_time: 0.0131 lr: 0.000000 max_mem: 10044M [08/07 15:25:26 d2.utils.events]: eta: 0:14:03 iter: 11319 total_loss: 0.153 loss_cls: 0.076 loss_box_reg: 0.076 time: 0.2287 data_time: 0.0139 lr: 0.000000 max_mem: 10044M [08/07 15:25:31 d2.utils.events]: eta: 0:13:57 iter: 11339 total_loss: 0.146 loss_cls: 0.072 loss_box_reg: 0.067 time: 0.2287 data_time: 0.0161 lr: 0.000000 max_mem: 10044M [08/07 15:25:35 d2.utils.events]: eta: 0:13:53 iter: 11359 total_loss: 0.139 loss_cls: 0.072 loss_box_reg: 0.066 time: 0.2287 data_time: 0.0143 lr: 0.000000 max_mem: 10044M [08/07 15:25:40 d2.utils.events]: eta: 0:13:49 iter: 11379 total_loss: 0.141 loss_cls: 0.072 loss_box_reg: 0.069 time: 0.2287 data_time: 0.0120 lr: 0.000000 max_mem: 10044M [08/07 15:25:45 d2.utils.events]: eta: 0:13:45 iter: 11399 total_loss: 0.150 loss_cls: 0.076 loss_box_reg: 0.070 time: 0.2287 data_time: 0.0149 lr: 0.000000 max_mem: 10044M [08/07 15:25:49 d2.utils.events]: eta: 0:13:40 iter: 11419 total_loss: 0.132 loss_cls: 0.066 loss_box_reg: 0.064 time: 0.2287 data_time: 0.0132 lr: 0.000000 max_mem: 10044M [08/07 15:25:53 d2.utils.events]: eta: 0:13:34 iter: 11439 total_loss: 0.142 loss_cls: 0.071 loss_box_reg: 0.065 time: 0.2287 data_time: 0.0118 lr: 0.000000 max_mem: 10044M [08/07 15:25:58 d2.utils.events]: eta: 0:13:30 iter: 11459 total_loss: 0.147 loss_cls: 0.074 loss_box_reg: 0.070 time: 0.2287 data_time: 0.0143 lr: 0.000000 max_mem: 10044M [08/07 15:26:03 d2.utils.events]: eta: 0:13:25 iter: 11479 total_loss: 0.138 loss_cls: 0.072 loss_box_reg: 0.068 time: 0.2287 data_time: 0.0132 lr: 0.000000 max_mem: 10044M [08/07 15:26:07 d2.utils.events]: eta: 0:13:19 iter: 11499 total_loss: 0.138 loss_cls: 0.072 loss_box_reg: 0.064 time: 0.2286 data_time: 0.0123 lr: 0.000000 max_mem: 10044M [08/07 15:26:12 d2.utils.events]: eta: 0:13:14 iter: 11519 total_loss: 0.156 loss_cls: 0.077 loss_box_reg: 0.075 time: 0.2286 data_time: 0.0153 lr: 0.000000 max_mem: 10044M [08/07 15:26:16 d2.utils.events]: eta: 0:13:10 iter: 11539 total_loss: 0.130 loss_cls: 0.067 loss_box_reg: 0.061 time: 0.2286 data_time: 0.0114 lr: 0.000000 max_mem: 10044M [08/07 15:26:21 d2.utils.events]: eta: 0:13:05 iter: 11559 total_loss: 0.137 loss_cls: 0.072 loss_box_reg: 0.066 time: 0.2286 data_time: 0.0137 lr: 0.000000 max_mem: 10044M [08/07 15:26:25 d2.utils.events]: eta: 0:13:01 iter: 11579 total_loss: 0.144 loss_cls: 0.074 loss_box_reg: 0.071 time: 0.2286 data_time: 0.0133 lr: 0.000000 max_mem: 10044M [08/07 15:26:30 d2.utils.events]: eta: 0:12:57 iter: 11599 total_loss: 0.140 loss_cls: 0.073 loss_box_reg: 0.067 time: 0.2286 data_time: 0.0111 lr: 0.000000 max_mem: 10044M [08/07 15:26:34 d2.utils.events]: eta: 0:12:52 iter: 11619 total_loss: 0.129 loss_cls: 0.068 loss_box_reg: 0.061 time: 0.2286 data_time: 0.0129 lr: 0.000000 max_mem: 10044M [08/07 15:26:39 d2.utils.events]: eta: 0:12:47 iter: 11639 total_loss: 0.147 loss_cls: 0.074 loss_box_reg: 0.069 time: 0.2286 data_time: 0.0126 lr: 0.000000 max_mem: 10044M [08/07 15:26:43 d2.utils.events]: eta: 0:12:41 iter: 11659 total_loss: 0.134 loss_cls: 0.069 loss_box_reg: 0.064 time: 0.2286 data_time: 0.0132 lr: 0.000000 max_mem: 10044M [08/07 15:26:48 d2.utils.events]: eta: 0:12:36 iter: 11679 total_loss: 0.151 loss_cls: 0.078 loss_box_reg: 0.077 time: 0.2286 data_time: 0.0124 lr: 0.000000 max_mem: 10044M [08/07 15:26:52 d2.utils.events]: eta: 0:12:32 iter: 11699 total_loss: 0.143 loss_cls: 0.076 loss_box_reg: 0.067 time: 0.2285 data_time: 0.0128 lr: 0.000000 max_mem: 10044M [08/07 15:26:57 d2.utils.events]: eta: 0:12:27 iter: 11719 total_loss: 0.125 loss_cls: 0.068 loss_box_reg: 0.059 time: 0.2285 data_time: 0.0108 lr: 0.000000 max_mem: 10044M [08/07 15:27:01 d2.utils.events]: eta: 0:12:22 iter: 11739 total_loss: 0.141 loss_cls: 0.073 loss_box_reg: 0.069 time: 0.2285 data_time: 0.0129 lr: 0.000000 max_mem: 10044M [08/07 15:27:06 d2.utils.events]: eta: 0:12:18 iter: 11759 total_loss: 0.148 loss_cls: 0.079 loss_box_reg: 0.067 time: 0.2285 data_time: 0.0120 lr: 0.000000 max_mem: 10044M [08/07 15:27:10 d2.utils.events]: eta: 0:12:13 iter: 11779 total_loss: 0.129 loss_cls: 0.070 loss_box_reg: 0.064 time: 0.2285 data_time: 0.0118 lr: 0.000000 max_mem: 10044M [08/07 15:27:14 d2.utils.events]: eta: 0:12:08 iter: 11799 total_loss: 0.141 loss_cls: 0.073 loss_box_reg: 0.067 time: 0.2284 data_time: 0.0132 lr: 0.000000 max_mem: 10044M [08/07 15:27:19 d2.utils.events]: eta: 0:12:03 iter: 11819 total_loss: 0.132 loss_cls: 0.070 loss_box_reg: 0.062 time: 0.2284 data_time: 0.0129 lr: 0.000000 max_mem: 10044M [08/07 15:27:24 d2.utils.events]: eta: 0:11:59 iter: 11839 total_loss: 0.136 loss_cls: 0.070 loss_box_reg: 0.068 time: 0.2284 data_time: 0.0143 lr: 0.000000 max_mem: 10044M [08/07 15:27:28 d2.utils.events]: eta: 0:11:54 iter: 11859 total_loss: 0.134 loss_cls: 0.073 loss_box_reg: 0.063 time: 0.2284 data_time: 0.0111 lr: 0.000000 max_mem: 10044M [08/07 15:27:33 d2.utils.events]: eta: 0:11:50 iter: 11879 total_loss: 0.138 loss_cls: 0.072 loss_box_reg: 0.069 time: 0.2284 data_time: 0.0121 lr: 0.000000 max_mem: 10044M [08/07 15:27:38 d2.utils.events]: eta: 0:11:46 iter: 11899 total_loss: 0.139 loss_cls: 0.071 loss_box_reg: 0.066 time: 0.2285 data_time: 0.0122 lr: 0.000000 max_mem: 10044M [08/07 15:27:43 d2.utils.events]: eta: 0:11:43 iter: 11919 total_loss: 0.137 loss_cls: 0.070 loss_box_reg: 0.070 time: 0.2285 data_time: 0.0142 lr: 0.000000 max_mem: 10044M [08/07 15:27:48 d2.utils.events]: eta: 0:11:38 iter: 11939 total_loss: 0.137 loss_cls: 0.070 loss_box_reg: 0.065 time: 0.2286 data_time: 0.0127 lr: 0.000000 max_mem: 10044M [08/07 15:27:52 d2.utils.events]: eta: 0:11:34 iter: 11959 total_loss: 0.133 loss_cls: 0.070 loss_box_reg: 0.063 time: 0.2286 data_time: 0.0139 lr: 0.000000 max_mem: 10044M [08/07 15:27:57 d2.utils.events]: eta: 0:11:30 iter: 11979 total_loss: 0.146 loss_cls: 0.077 loss_box_reg: 0.072 time: 0.2286 data_time: 0.0144 lr: 0.000000 max_mem: 10044M [08/07 15:28:02 d2.utils.events]: eta: 0:11:26 iter: 11999 total_loss: 0.135 loss_cls: 0.068 loss_box_reg: 0.061 time: 0.2287 data_time: 0.0141 lr: 0.000000 max_mem: 10044M [08/07 15:28:07 d2.utils.events]: eta: 0:11:23 iter: 12019 total_loss: 0.142 loss_cls: 0.071 loss_box_reg: 0.071 time: 0.2288 data_time: 0.0161 lr: 0.000000 max_mem: 10044M [08/07 15:28:12 d2.utils.events]: eta: 0:11:19 iter: 12039 total_loss: 0.131 loss_cls: 0.070 loss_box_reg: 0.064 time: 0.2288 data_time: 0.0140 lr: 0.000000 max_mem: 10044M [08/07 15:28:17 d2.utils.events]: eta: 0:11:14 iter: 12059 total_loss: 0.144 loss_cls: 0.072 loss_box_reg: 0.070 time: 0.2288 data_time: 0.0115 lr: 0.000000 max_mem: 10044M [08/07 15:28:22 d2.utils.events]: eta: 0:11:11 iter: 12079 total_loss: 0.132 loss_cls: 0.069 loss_box_reg: 0.065 time: 0.2289 data_time: 0.0188 lr: 0.000000 max_mem: 10044M [08/07 15:28:27 d2.utils.events]: eta: 0:11:07 iter: 12099 total_loss: 0.150 loss_cls: 0.077 loss_box_reg: 0.073 time: 0.2289 data_time: 0.0134 lr: 0.000000 max_mem: 10044M [08/07 15:28:32 d2.utils.events]: eta: 0:11:02 iter: 12119 total_loss: 0.131 loss_cls: 0.070 loss_box_reg: 0.063 time: 0.2290 data_time: 0.0146 lr: 0.000000 max_mem: 10044M [08/07 15:28:36 d2.utils.events]: eta: 0:10:58 iter: 12139 total_loss: 0.135 loss_cls: 0.068 loss_box_reg: 0.068 time: 0.2290 data_time: 0.0148 lr: 0.000000 max_mem: 10044M [08/07 15:28:41 d2.utils.events]: eta: 0:10:53 iter: 12159 total_loss: 0.136 loss_cls: 0.073 loss_box_reg: 0.062 time: 0.2290 data_time: 0.0128 lr: 0.000000 max_mem: 10044M [08/07 15:28:46 d2.utils.events]: eta: 0:10:49 iter: 12179 total_loss: 0.146 loss_cls: 0.068 loss_box_reg: 0.075 time: 0.2290 data_time: 0.0133 lr: 0.000000 max_mem: 10044M [08/07 15:28:50 d2.utils.events]: eta: 0:10:44 iter: 12199 total_loss: 0.139 loss_cls: 0.073 loss_box_reg: 0.068 time: 0.2290 data_time: 0.0139 lr: 0.000000 max_mem: 10044M [08/07 15:28:55 d2.utils.events]: eta: 0:10:39 iter: 12219 total_loss: 0.142 loss_cls: 0.075 loss_box_reg: 0.071 time: 0.2290 data_time: 0.0114 lr: 0.000000 max_mem: 10044M [08/07 15:29:00 d2.utils.events]: eta: 0:10:35 iter: 12239 total_loss: 0.137 loss_cls: 0.069 loss_box_reg: 0.066 time: 0.2290 data_time: 0.0170 lr: 0.000000 max_mem: 10044M [08/07 15:29:04 d2.utils.events]: eta: 0:10:31 iter: 12259 total_loss: 0.150 loss_cls: 0.073 loss_box_reg: 0.075 time: 0.2290 data_time: 0.0130 lr: 0.000000 max_mem: 10044M [08/07 15:29:09 d2.utils.events]: eta: 0:10:26 iter: 12279 total_loss: 0.143 loss_cls: 0.072 loss_box_reg: 0.071 time: 0.2290 data_time: 0.0125 lr: 0.000000 max_mem: 10044M [08/07 15:29:13 d2.utils.events]: eta: 0:10:22 iter: 12299 total_loss: 0.137 loss_cls: 0.068 loss_box_reg: 0.071 time: 0.2290 data_time: 0.0171 lr: 0.000000 max_mem: 10044M [08/07 15:29:18 d2.utils.events]: eta: 0:10:17 iter: 12319 total_loss: 0.147 loss_cls: 0.073 loss_box_reg: 0.073 time: 0.2290 data_time: 0.0146 lr: 0.000000 max_mem: 10044M [08/07 15:29:23 d2.utils.events]: eta: 0:10:13 iter: 12339 total_loss: 0.138 loss_cls: 0.069 loss_box_reg: 0.068 time: 0.2290 data_time: 0.0122 lr: 0.000000 max_mem: 10044M [08/07 15:29:27 d2.utils.events]: eta: 0:10:08 iter: 12359 total_loss: 0.123 loss_cls: 0.065 loss_box_reg: 0.062 time: 0.2290 data_time: 0.0124 lr: 0.000000 max_mem: 10044M [08/07 15:29:32 d2.utils.events]: eta: 0:10:03 iter: 12379 total_loss: 0.139 loss_cls: 0.070 loss_box_reg: 0.070 time: 0.2290 data_time: 0.0118 lr: 0.000000 max_mem: 10044M [08/07 15:29:36 d2.utils.events]: eta: 0:09:58 iter: 12399 total_loss: 0.126 loss_cls: 0.065 loss_box_reg: 0.063 time: 0.2290 data_time: 0.0127 lr: 0.000000 max_mem: 10044M [08/07 15:29:41 d2.utils.events]: eta: 0:09:54 iter: 12419 total_loss: 0.153 loss_cls: 0.076 loss_box_reg: 0.073 time: 0.2290 data_time: 0.0114 lr: 0.000000 max_mem: 10044M [08/07 15:29:45 d2.utils.events]: eta: 0:09:50 iter: 12439 total_loss: 0.126 loss_cls: 0.065 loss_box_reg: 0.063 time: 0.2290 data_time: 0.0122 lr: 0.000000 max_mem: 10044M [08/07 15:29:50 d2.utils.events]: eta: 0:09:45 iter: 12459 total_loss: 0.134 loss_cls: 0.073 loss_box_reg: 0.063 time: 0.2290 data_time: 0.0138 lr: 0.000000 max_mem: 10044M [08/07 15:29:55 d2.utils.events]: eta: 0:09:41 iter: 12479 total_loss: 0.143 loss_cls: 0.070 loss_box_reg: 0.070 time: 0.2290 data_time: 0.0125 lr: 0.000000 max_mem: 10044M [08/07 15:29:59 d2.utils.events]: eta: 0:09:37 iter: 12499 total_loss: 0.147 loss_cls: 0.076 loss_box_reg: 0.071 time: 0.2290 data_time: 0.0108 lr: 0.000000 max_mem: 10044M [08/07 15:30:04 d2.utils.events]: eta: 0:09:33 iter: 12519 total_loss: 0.138 loss_cls: 0.071 loss_box_reg: 0.071 time: 0.2290 data_time: 0.0135 lr: 0.000000 max_mem: 10044M [08/07 15:30:08 d2.utils.events]: eta: 0:09:28 iter: 12539 total_loss: 0.136 loss_cls: 0.069 loss_box_reg: 0.063 time: 0.2290 data_time: 0.0105 lr: 0.000000 max_mem: 10044M [08/07 15:30:13 d2.utils.events]: eta: 0:09:24 iter: 12559 total_loss: 0.150 loss_cls: 0.074 loss_box_reg: 0.073 time: 0.2290 data_time: 0.0121 lr: 0.000000 max_mem: 10044M [08/07 15:30:17 d2.utils.events]: eta: 0:09:18 iter: 12579 total_loss: 0.134 loss_cls: 0.070 loss_box_reg: 0.064 time: 0.2290 data_time: 0.0128 lr: 0.000000 max_mem: 10044M [08/07 15:30:22 d2.utils.events]: eta: 0:09:14 iter: 12599 total_loss: 0.135 loss_cls: 0.072 loss_box_reg: 0.062 time: 0.2290 data_time: 0.0148 lr: 0.000000 max_mem: 10044M [08/07 15:30:27 d2.utils.events]: eta: 0:09:10 iter: 12619 total_loss: 0.137 loss_cls: 0.070 loss_box_reg: 0.067 time: 0.2290 data_time: 0.0115 lr: 0.000000 max_mem: 10044M [08/07 15:30:31 d2.utils.events]: eta: 0:09:07 iter: 12639 total_loss: 0.136 loss_cls: 0.073 loss_box_reg: 0.065 time: 0.2290 data_time: 0.0130 lr: 0.000000 max_mem: 10044M [08/07 15:30:36 d2.utils.events]: eta: 0:09:03 iter: 12659 total_loss: 0.145 loss_cls: 0.072 loss_box_reg: 0.069 time: 0.2290 data_time: 0.0148 lr: 0.000000 max_mem: 10044M [08/07 15:30:41 d2.utils.events]: eta: 0:09:00 iter: 12679 total_loss: 0.145 loss_cls: 0.074 loss_box_reg: 0.077 time: 0.2291 data_time: 0.0104 lr: 0.000000 max_mem: 10044M [08/07 15:30:46 d2.utils.events]: eta: 0:08:57 iter: 12699 total_loss: 0.124 loss_cls: 0.064 loss_box_reg: 0.061 time: 0.2291 data_time: 0.0125 lr: 0.000000 max_mem: 10044M [08/07 15:30:51 d2.utils.events]: eta: 0:08:54 iter: 12719 total_loss: 0.124 loss_cls: 0.063 loss_box_reg: 0.063 time: 0.2291 data_time: 0.0145 lr: 0.000000 max_mem: 10044M [08/07 15:30:56 d2.utils.events]: eta: 0:08:49 iter: 12739 total_loss: 0.152 loss_cls: 0.080 loss_box_reg: 0.072 time: 0.2292 data_time: 0.0132 lr: 0.000000 max_mem: 10044M [08/07 15:31:01 d2.utils.events]: eta: 0:08:44 iter: 12759 total_loss: 0.144 loss_cls: 0.072 loss_box_reg: 0.073 time: 0.2292 data_time: 0.0112 lr: 0.000000 max_mem: 10044M [08/07 15:31:05 d2.utils.events]: eta: 0:08:41 iter: 12779 total_loss: 0.124 loss_cls: 0.064 loss_box_reg: 0.060 time: 0.2292 data_time: 0.0143 lr: 0.000000 max_mem: 10044M [08/07 15:31:10 d2.utils.events]: eta: 0:08:37 iter: 12799 total_loss: 0.139 loss_cls: 0.072 loss_box_reg: 0.066 time: 0.2292 data_time: 0.0167 lr: 0.000000 max_mem: 10044M [08/07 15:31:15 d2.utils.events]: eta: 0:08:34 iter: 12819 total_loss: 0.134 loss_cls: 0.069 loss_box_reg: 0.066 time: 0.2293 data_time: 0.0167 lr: 0.000000 max_mem: 10044M [08/07 15:31:20 d2.utils.events]: eta: 0:08:29 iter: 12839 total_loss: 0.138 loss_cls: 0.073 loss_box_reg: 0.066 time: 0.2293 data_time: 0.0133 lr: 0.000000 max_mem: 10044M [08/07 15:31:25 d2.utils.events]: eta: 0:08:27 iter: 12859 total_loss: 0.137 loss_cls: 0.072 loss_box_reg: 0.066 time: 0.2293 data_time: 0.0120 lr: 0.000000 max_mem: 10044M [08/07 15:31:30 d2.utils.events]: eta: 0:08:23 iter: 12879 total_loss: 0.127 loss_cls: 0.067 loss_box_reg: 0.063 time: 0.2294 data_time: 0.0147 lr: 0.000000 max_mem: 10044M [08/07 15:31:34 d2.utils.events]: eta: 0:08:17 iter: 12899 total_loss: 0.144 loss_cls: 0.074 loss_box_reg: 0.067 time: 0.2294 data_time: 0.0114 lr: 0.000000 max_mem: 10044M [08/07 15:31:39 d2.utils.events]: eta: 0:08:11 iter: 12919 total_loss: 0.131 loss_cls: 0.070 loss_box_reg: 0.064 time: 0.2294 data_time: 0.0132 lr: 0.000000 max_mem: 10044M [08/07 15:31:43 d2.utils.events]: eta: 0:08:07 iter: 12939 total_loss: 0.128 loss_cls: 0.069 loss_box_reg: 0.060 time: 0.2294 data_time: 0.0127 lr: 0.000000 max_mem: 10044M [08/07 15:31:48 d2.utils.events]: eta: 0:08:01 iter: 12959 total_loss: 0.137 loss_cls: 0.073 loss_box_reg: 0.061 time: 0.2294 data_time: 0.0122 lr: 0.000000 max_mem: 10044M [08/07 15:31:52 d2.utils.events]: eta: 0:07:55 iter: 12979 total_loss: 0.137 loss_cls: 0.071 loss_box_reg: 0.068 time: 0.2293 data_time: 0.0127 lr: 0.000000 max_mem: 10044M [08/07 15:31:57 d2.utils.events]: eta: 0:07:51 iter: 12999 total_loss: 0.156 loss_cls: 0.084 loss_box_reg: 0.075 time: 0.2293 data_time: 0.0127 lr: 0.000000 max_mem: 10044M [08/07 15:32:01 d2.utils.events]: eta: 0:07:45 iter: 13019 total_loss: 0.154 loss_cls: 0.075 loss_box_reg: 0.071 time: 0.2293 data_time: 0.0137 lr: 0.000000 max_mem: 10044M [08/07 15:32:06 d2.utils.events]: eta: 0:07:39 iter: 13039 total_loss: 0.127 loss_cls: 0.066 loss_box_reg: 0.059 time: 0.2293 data_time: 0.0108 lr: 0.000000 max_mem: 10044M [08/07 15:32:11 d2.utils.events]: eta: 0:07:34 iter: 13059 total_loss: 0.146 loss_cls: 0.072 loss_box_reg: 0.071 time: 0.2293 data_time: 0.0125 lr: 0.000000 max_mem: 10044M [08/07 15:32:15 d2.utils.events]: eta: 0:07:28 iter: 13079 total_loss: 0.119 loss_cls: 0.064 loss_box_reg: 0.057 time: 0.2293 data_time: 0.0101 lr: 0.000000 max_mem: 10044M [08/07 15:32:20 d2.utils.events]: eta: 0:07:22 iter: 13099 total_loss: 0.130 loss_cls: 0.070 loss_box_reg: 0.063 time: 0.2293 data_time: 0.0143 lr: 0.000000 max_mem: 10044M [08/07 15:32:24 d2.utils.events]: eta: 0:07:16 iter: 13119 total_loss: 0.131 loss_cls: 0.067 loss_box_reg: 0.065 time: 0.2293 data_time: 0.0128 lr: 0.000000 max_mem: 10044M [08/07 15:32:29 d2.utils.events]: eta: 0:07:11 iter: 13139 total_loss: 0.136 loss_cls: 0.071 loss_box_reg: 0.065 time: 0.2293 data_time: 0.0131 lr: 0.000000 max_mem: 10044M [08/07 15:32:34 d2.utils.events]: eta: 0:07:06 iter: 13159 total_loss: 0.119 loss_cls: 0.066 loss_box_reg: 0.058 time: 0.2293 data_time: 0.0127 lr: 0.000000 max_mem: 10044M [08/07 15:32:38 d2.utils.events]: eta: 0:07:02 iter: 13179 total_loss: 0.131 loss_cls: 0.069 loss_box_reg: 0.062 time: 0.2293 data_time: 0.0163 lr: 0.000000 max_mem: 10044M [08/07 15:32:43 d2.utils.events]: eta: 0:06:57 iter: 13199 total_loss: 0.144 loss_cls: 0.077 loss_box_reg: 0.071 time: 0.2293 data_time: 0.0128 lr: 0.000000 max_mem: 10044M [08/07 15:32:47 d2.utils.events]: eta: 0:06:52 iter: 13219 total_loss: 0.132 loss_cls: 0.068 loss_box_reg: 0.065 time: 0.2293 data_time: 0.0142 lr: 0.000000 max_mem: 10044M [08/07 15:32:52 d2.utils.events]: eta: 0:06:48 iter: 13239 total_loss: 0.132 loss_cls: 0.071 loss_box_reg: 0.059 time: 0.2293 data_time: 0.0126 lr: 0.000000 max_mem: 10044M [08/07 15:32:57 d2.utils.events]: eta: 0:06:43 iter: 13259 total_loss: 0.135 loss_cls: 0.070 loss_box_reg: 0.066 time: 0.2293 data_time: 0.0135 lr: 0.000000 max_mem: 10044M [08/07 15:33:01 d2.utils.events]: eta: 0:06:38 iter: 13279 total_loss: 0.144 loss_cls: 0.072 loss_box_reg: 0.072 time: 0.2293 data_time: 0.0146 lr: 0.000000 max_mem: 10044M [08/07 15:33:06 d2.utils.events]: eta: 0:06:33 iter: 13299 total_loss: 0.130 loss_cls: 0.068 loss_box_reg: 0.065 time: 0.2293 data_time: 0.0147 lr: 0.000000 max_mem: 10044M [08/07 15:33:11 d2.utils.events]: eta: 0:06:29 iter: 13319 total_loss: 0.136 loss_cls: 0.066 loss_box_reg: 0.066 time: 0.2293 data_time: 0.0103 lr: 0.000000 max_mem: 10044M [08/07 15:33:15 d2.utils.events]: eta: 0:06:24 iter: 13339 total_loss: 0.135 loss_cls: 0.070 loss_box_reg: 0.064 time: 0.2293 data_time: 0.0129 lr: 0.000000 max_mem: 10044M [08/07 15:33:20 d2.utils.events]: eta: 0:06:20 iter: 13359 total_loss: 0.148 loss_cls: 0.077 loss_box_reg: 0.071 time: 0.2293 data_time: 0.0103 lr: 0.000000 max_mem: 10044M [08/07 15:33:24 d2.utils.events]: eta: 0:06:15 iter: 13379 total_loss: 0.146 loss_cls: 0.075 loss_box_reg: 0.070 time: 0.2293 data_time: 0.0151 lr: 0.000000 max_mem: 10044M [08/07 15:33:29 d2.utils.events]: eta: 0:06:11 iter: 13399 total_loss: 0.154 loss_cls: 0.079 loss_box_reg: 0.073 time: 0.2293 data_time: 0.0145 lr: 0.000000 max_mem: 10044M [08/07 15:33:33 d2.utils.events]: eta: 0:06:06 iter: 13419 total_loss: 0.133 loss_cls: 0.066 loss_box_reg: 0.062 time: 0.2293 data_time: 0.0129 lr: 0.000000 max_mem: 10044M [08/07 15:33:38 d2.utils.events]: eta: 0:06:01 iter: 13439 total_loss: 0.150 loss_cls: 0.080 loss_box_reg: 0.070 time: 0.2293 data_time: 0.0103 lr: 0.000000 max_mem: 10044M [08/07 15:33:42 d2.utils.events]: eta: 0:05:57 iter: 13459 total_loss: 0.141 loss_cls: 0.072 loss_box_reg: 0.071 time: 0.2293 data_time: 0.0125 lr: 0.000000 max_mem: 10044M [08/07 15:33:47 d2.utils.events]: eta: 0:05:53 iter: 13479 total_loss: 0.148 loss_cls: 0.073 loss_box_reg: 0.074 time: 0.2293 data_time: 0.0135 lr: 0.000000 max_mem: 10044M [08/07 15:33:52 d2.utils.events]: eta: 0:05:48 iter: 13499 total_loss: 0.137 loss_cls: 0.071 loss_box_reg: 0.065 time: 0.2293 data_time: 0.0116 lr: 0.000000 max_mem: 10044M [08/07 15:33:56 d2.utils.events]: eta: 0:05:43 iter: 13519 total_loss: 0.126 loss_cls: 0.066 loss_box_reg: 0.061 time: 0.2293 data_time: 0.0138 lr: 0.000000 max_mem: 10044M [08/07 15:34:01 d2.utils.events]: eta: 0:05:39 iter: 13539 total_loss: 0.129 loss_cls: 0.068 loss_box_reg: 0.060 time: 0.2293 data_time: 0.0127 lr: 0.000000 max_mem: 10044M [08/07 15:34:06 d2.utils.events]: eta: 0:05:34 iter: 13559 total_loss: 0.128 loss_cls: 0.065 loss_box_reg: 0.061 time: 0.2293 data_time: 0.0135 lr: 0.000000 max_mem: 10044M [08/07 15:34:10 d2.utils.events]: eta: 0:05:30 iter: 13579 total_loss: 0.145 loss_cls: 0.075 loss_box_reg: 0.070 time: 0.2293 data_time: 0.0121 lr: 0.000000 max_mem: 10044M [08/07 15:34:15 d2.utils.events]: eta: 0:05:26 iter: 13599 total_loss: 0.140 loss_cls: 0.073 loss_box_reg: 0.065 time: 0.2293 data_time: 0.0115 lr: 0.000000 max_mem: 10044M [08/07 15:34:20 d2.utils.events]: eta: 0:05:20 iter: 13619 total_loss: 0.154 loss_cls: 0.081 loss_box_reg: 0.072 time: 0.2293 data_time: 0.0129 lr: 0.000000 max_mem: 10044M [08/07 15:34:24 d2.utils.events]: eta: 0:05:15 iter: 13639 total_loss: 0.139 loss_cls: 0.072 loss_box_reg: 0.072 time: 0.2293 data_time: 0.0167 lr: 0.000000 max_mem: 10044M [08/07 15:34:29 d2.utils.events]: eta: 0:05:11 iter: 13659 total_loss: 0.135 loss_cls: 0.067 loss_box_reg: 0.065 time: 0.2293 data_time: 0.0137 lr: 0.000000 max_mem: 10044M [08/07 15:34:33 d2.utils.events]: eta: 0:05:06 iter: 13679 total_loss: 0.147 loss_cls: 0.073 loss_box_reg: 0.071 time: 0.2293 data_time: 0.0154 lr: 0.000000 max_mem: 10044M [08/07 15:34:38 d2.utils.events]: eta: 0:05:01 iter: 13699 total_loss: 0.148 loss_cls: 0.074 loss_box_reg: 0.069 time: 0.2293 data_time: 0.0132 lr: 0.000000 max_mem: 10044M [08/07 15:34:43 d2.utils.events]: eta: 0:04:56 iter: 13719 total_loss: 0.141 loss_cls: 0.071 loss_box_reg: 0.065 time: 0.2293 data_time: 0.0150 lr: 0.000000 max_mem: 10044M [08/07 15:34:47 d2.utils.events]: eta: 0:04:51 iter: 13739 total_loss: 0.136 loss_cls: 0.067 loss_box_reg: 0.067 time: 0.2293 data_time: 0.0136 lr: 0.000000 max_mem: 10044M [08/07 15:34:52 d2.utils.events]: eta: 0:04:46 iter: 13759 total_loss: 0.139 loss_cls: 0.073 loss_box_reg: 0.068 time: 0.2293 data_time: 0.0126 lr: 0.000000 max_mem: 10044M [08/07 15:34:56 d2.utils.events]: eta: 0:04:41 iter: 13779 total_loss: 0.128 loss_cls: 0.066 loss_box_reg: 0.061 time: 0.2293 data_time: 0.0151 lr: 0.000000 max_mem: 10044M [08/07 15:35:01 d2.utils.events]: eta: 0:04:36 iter: 13799 total_loss: 0.134 loss_cls: 0.073 loss_box_reg: 0.065 time: 0.2293 data_time: 0.0136 lr: 0.000000 max_mem: 10044M [08/07 15:35:06 d2.utils.events]: eta: 0:04:31 iter: 13819 total_loss: 0.136 loss_cls: 0.069 loss_box_reg: 0.067 time: 0.2293 data_time: 0.0127 lr: 0.000000 max_mem: 10044M [08/07 15:35:10 d2.utils.events]: eta: 0:04:26 iter: 13839 total_loss: 0.145 loss_cls: 0.072 loss_box_reg: 0.070 time: 0.2293 data_time: 0.0124 lr: 0.000000 max_mem: 10044M [08/07 15:35:15 d2.utils.events]: eta: 0:04:22 iter: 13859 total_loss: 0.129 loss_cls: 0.067 loss_box_reg: 0.064 time: 0.2293 data_time: 0.0137 lr: 0.000000 max_mem: 10044M [08/07 15:35:20 d2.utils.events]: eta: 0:04:16 iter: 13879 total_loss: 0.152 loss_cls: 0.076 loss_box_reg: 0.073 time: 0.2293 data_time: 0.0100 lr: 0.000000 max_mem: 10044M [08/07 15:35:24 d2.utils.events]: eta: 0:04:12 iter: 13899 total_loss: 0.159 loss_cls: 0.079 loss_box_reg: 0.074 time: 0.2294 data_time: 0.0133 lr: 0.000000 max_mem: 10044M [08/07 15:35:29 d2.utils.events]: eta: 0:04:07 iter: 13919 total_loss: 0.140 loss_cls: 0.073 loss_box_reg: 0.069 time: 0.2294 data_time: 0.0154 lr: 0.000000 max_mem: 10044M [08/07 15:35:34 d2.utils.events]: eta: 0:04:03 iter: 13939 total_loss: 0.153 loss_cls: 0.075 loss_box_reg: 0.078 time: 0.2294 data_time: 0.0129 lr: 0.000000 max_mem: 10044M [08/07 15:35:38 d2.utils.events]: eta: 0:03:59 iter: 13959 total_loss: 0.135 loss_cls: 0.069 loss_box_reg: 0.067 time: 0.2294 data_time: 0.0154 lr: 0.000000 max_mem: 10044M [08/07 15:35:43 d2.utils.events]: eta: 0:03:54 iter: 13979 total_loss: 0.134 loss_cls: 0.068 loss_box_reg: 0.068 time: 0.2294 data_time: 0.0145 lr: 0.000000 max_mem: 10044M [08/07 15:35:48 d2.utils.events]: eta: 0:03:50 iter: 13999 total_loss: 0.138 loss_cls: 0.071 loss_box_reg: 0.067 time: 0.2294 data_time: 0.0137 lr: 0.000000 max_mem: 10044M [08/07 15:35:53 d2.utils.events]: eta: 0:03:46 iter: 14019 total_loss: 0.131 loss_cls: 0.068 loss_box_reg: 0.063 time: 0.2294 data_time: 0.0141 lr: 0.000000 max_mem: 10044M [08/07 15:35:57 d2.utils.events]: eta: 0:03:41 iter: 14039 total_loss: 0.131 loss_cls: 0.070 loss_box_reg: 0.059 time: 0.2295 data_time: 0.0154 lr: 0.000000 max_mem: 10044M [08/07 15:36:02 d2.utils.events]: eta: 0:03:37 iter: 14059 total_loss: 0.142 loss_cls: 0.070 loss_box_reg: 0.067 time: 0.2295 data_time: 0.0124 lr: 0.000000 max_mem: 10044M [08/07 15:36:07 d2.utils.events]: eta: 0:03:32 iter: 14079 total_loss: 0.134 loss_cls: 0.071 loss_box_reg: 0.066 time: 0.2295 data_time: 0.0141 lr: 0.000000 max_mem: 10044M [08/07 15:36:11 d2.utils.events]: eta: 0:03:28 iter: 14099 total_loss: 0.145 loss_cls: 0.070 loss_box_reg: 0.074 time: 0.2295 data_time: 0.0133 lr: 0.000000 max_mem: 10044M [08/07 15:36:16 d2.utils.events]: eta: 0:03:23 iter: 14119 total_loss: 0.144 loss_cls: 0.072 loss_box_reg: 0.069 time: 0.2295 data_time: 0.0112 lr: 0.000000 max_mem: 10044M [08/07 15:36:20 d2.utils.events]: eta: 0:03:18 iter: 14139 total_loss: 0.144 loss_cls: 0.074 loss_box_reg: 0.070 time: 0.2294 data_time: 0.0113 lr: 0.000000 max_mem: 10044M [08/07 15:36:25 d2.utils.events]: eta: 0:03:14 iter: 14159 total_loss: 0.143 loss_cls: 0.071 loss_box_reg: 0.070 time: 0.2295 data_time: 0.0115 lr: 0.000000 max_mem: 10044M [08/07 15:36:30 d2.utils.events]: eta: 0:03:09 iter: 14179 total_loss: 0.123 loss_cls: 0.066 loss_box_reg: 0.059 time: 0.2294 data_time: 0.0160 lr: 0.000000 max_mem: 10044M [08/07 15:36:34 d2.utils.events]: eta: 0:03:05 iter: 14199 total_loss: 0.147 loss_cls: 0.076 loss_box_reg: 0.073 time: 0.2294 data_time: 0.0126 lr: 0.000000 max_mem: 10044M [08/07 15:36:39 d2.utils.events]: eta: 0:03:00 iter: 14219 total_loss: 0.161 loss_cls: 0.081 loss_box_reg: 0.076 time: 0.2294 data_time: 0.0113 lr: 0.000000 max_mem: 10044M [08/07 15:36:43 d2.utils.events]: eta: 0:02:55 iter: 14239 total_loss: 0.144 loss_cls: 0.076 loss_box_reg: 0.071 time: 0.2294 data_time: 0.0121 lr: 0.000000 max_mem: 10044M [08/07 15:36:48 d2.utils.events]: eta: 0:02:50 iter: 14259 total_loss: 0.141 loss_cls: 0.074 loss_box_reg: 0.068 time: 0.2294 data_time: 0.0113 lr: 0.000000 max_mem: 10044M [08/07 15:36:52 d2.utils.events]: eta: 0:02:45 iter: 14279 total_loss: 0.136 loss_cls: 0.067 loss_box_reg: 0.065 time: 0.2294 data_time: 0.0161 lr: 0.000000 max_mem: 10044M [08/07 15:36:56 d2.utils.events]: eta: 0:02:41 iter: 14299 total_loss: 0.150 loss_cls: 0.078 loss_box_reg: 0.071 time: 0.2294 data_time: 0.0128 lr: 0.000000 max_mem: 10044M [08/07 15:37:01 d2.utils.events]: eta: 0:02:36 iter: 14319 total_loss: 0.131 loss_cls: 0.070 loss_box_reg: 0.065 time: 0.2293 data_time: 0.0127 lr: 0.000000 max_mem: 10044M [08/07 15:37:05 d2.utils.events]: eta: 0:02:31 iter: 14339 total_loss: 0.145 loss_cls: 0.072 loss_box_reg: 0.070 time: 0.2293 data_time: 0.0091 lr: 0.000000 max_mem: 10044M [08/07 15:37:10 d2.utils.events]: eta: 0:02:27 iter: 14359 total_loss: 0.140 loss_cls: 0.073 loss_box_reg: 0.068 time: 0.2293 data_time: 0.0104 lr: 0.000000 max_mem: 10044M [08/07 15:37:14 d2.utils.events]: eta: 0:02:22 iter: 14379 total_loss: 0.134 loss_cls: 0.069 loss_box_reg: 0.069 time: 0.2293 data_time: 0.0131 lr: 0.000000 max_mem: 10044M [08/07 15:37:18 d2.utils.events]: eta: 0:02:17 iter: 14399 total_loss: 0.133 loss_cls: 0.069 loss_box_reg: 0.065 time: 0.2292 data_time: 0.0119 lr: 0.000000 max_mem: 10044M [08/07 15:37:23 d2.utils.events]: eta: 0:02:12 iter: 14419 total_loss: 0.145 loss_cls: 0.075 loss_box_reg: 0.070 time: 0.2292 data_time: 0.0141 lr: 0.000000 max_mem: 10044M [08/07 15:37:27 d2.utils.events]: eta: 0:02:08 iter: 14439 total_loss: 0.145 loss_cls: 0.075 loss_box_reg: 0.071 time: 0.2292 data_time: 0.0119 lr: 0.000000 max_mem: 10044M [08/07 15:37:32 d2.utils.events]: eta: 0:02:03 iter: 14459 total_loss: 0.140 loss_cls: 0.073 loss_box_reg: 0.065 time: 0.2292 data_time: 0.0119 lr: 0.000000 max_mem: 10044M [08/07 15:37:36 d2.utils.events]: eta: 0:01:58 iter: 14479 total_loss: 0.137 loss_cls: 0.069 loss_box_reg: 0.067 time: 0.2292 data_time: 0.0114 lr: 0.000000 max_mem: 10044M [08/07 15:37:41 d2.utils.events]: eta: 0:01:54 iter: 14499 total_loss: 0.139 loss_cls: 0.070 loss_box_reg: 0.065 time: 0.2292 data_time: 0.0105 lr: 0.000000 max_mem: 10044M [08/07 15:37:45 d2.utils.events]: eta: 0:01:49 iter: 14519 total_loss: 0.133 loss_cls: 0.067 loss_box_reg: 0.066 time: 0.2292 data_time: 0.0137 lr: 0.000000 max_mem: 10044M [08/07 15:37:50 d2.utils.events]: eta: 0:01:45 iter: 14539 total_loss: 0.139 loss_cls: 0.069 loss_box_reg: 0.067 time: 0.2291 data_time: 0.0136 lr: 0.000000 max_mem: 10044M [08/07 15:37:54 d2.utils.events]: eta: 0:01:40 iter: 14559 total_loss: 0.153 loss_cls: 0.078 loss_box_reg: 0.076 time: 0.2291 data_time: 0.0122 lr: 0.000000 max_mem: 10044M [08/07 15:37:59 d2.utils.events]: eta: 0:01:35 iter: 14579 total_loss: 0.146 loss_cls: 0.071 loss_box_reg: 0.073 time: 0.2291 data_time: 0.0112 lr: 0.000000 max_mem: 10044M [08/07 15:38:03 d2.utils.events]: eta: 0:01:31 iter: 14599 total_loss: 0.131 loss_cls: 0.070 loss_box_reg: 0.066 time: 0.2291 data_time: 0.0138 lr: 0.000000 max_mem: 10044M [08/07 15:38:08 d2.utils.events]: eta: 0:01:26 iter: 14619 total_loss: 0.149 loss_cls: 0.075 loss_box_reg: 0.075 time: 0.2291 data_time: 0.0116 lr: 0.000000 max_mem: 10044M [08/07 15:38:12 d2.utils.events]: eta: 0:01:22 iter: 14639 total_loss: 0.139 loss_cls: 0.073 loss_box_reg: 0.066 time: 0.2291 data_time: 0.0105 lr: 0.000000 max_mem: 10044M [08/07 15:38:17 d2.utils.events]: eta: 0:01:17 iter: 14659 total_loss: 0.148 loss_cls: 0.074 loss_box_reg: 0.071 time: 0.2291 data_time: 0.0139 lr: 0.000000 max_mem: 10044M [08/07 15:38:22 d2.utils.events]: eta: 0:01:13 iter: 14679 total_loss: 0.120 loss_cls: 0.064 loss_box_reg: 0.059 time: 0.2291 data_time: 0.0119 lr: 0.000000 max_mem: 10044M [08/07 15:38:26 d2.utils.events]: eta: 0:01:08 iter: 14699 total_loss: 0.128 loss_cls: 0.066 loss_box_reg: 0.063 time: 0.2291 data_time: 0.0140 lr: 0.000000 max_mem: 10044M [08/07 15:38:31 d2.utils.events]: eta: 0:01:04 iter: 14719 total_loss: 0.134 loss_cls: 0.071 loss_box_reg: 0.066 time: 0.2292 data_time: 0.0146 lr: 0.000000 max_mem: 10044M [08/07 15:38:36 d2.utils.events]: eta: 0:00:59 iter: 14739 total_loss: 0.131 loss_cls: 0.068 loss_box_reg: 0.064 time: 0.2292 data_time: 0.0129 lr: 0.000000 max_mem: 10044M [08/07 15:38:40 d2.utils.events]: eta: 0:00:54 iter: 14759 total_loss: 0.134 loss_cls: 0.069 loss_box_reg: 0.065 time: 0.2291 data_time: 0.0124 lr: 0.000000 max_mem: 10044M [08/07 15:38:45 d2.utils.events]: eta: 0:00:50 iter: 14779 total_loss: 0.138 loss_cls: 0.069 loss_box_reg: 0.069 time: 0.2291 data_time: 0.0127 lr: 0.000000 max_mem: 10044M [08/07 15:38:50 d2.utils.events]: eta: 0:00:45 iter: 14799 total_loss: 0.128 loss_cls: 0.070 loss_box_reg: 0.061 time: 0.2292 data_time: 0.0123 lr: 0.000000 max_mem: 10044M [08/07 15:38:55 d2.utils.events]: eta: 0:00:41 iter: 14819 total_loss: 0.150 loss_cls: 0.078 loss_box_reg: 0.074 time: 0.2292 data_time: 0.0136 lr: 0.000000 max_mem: 10044M [08/07 15:39:00 d2.utils.events]: eta: 0:00:36 iter: 14839 total_loss: 0.128 loss_cls: 0.065 loss_box_reg: 0.063 time: 0.2292 data_time: 0.0146 lr: 0.000000 max_mem: 10044M [08/07 15:39:04 d2.utils.events]: eta: 0:00:32 iter: 14859 total_loss: 0.140 loss_cls: 0.069 loss_box_reg: 0.070 time: 0.2292 data_time: 0.0144 lr: 0.000000 max_mem: 10044M [08/07 15:39:09 d2.utils.events]: eta: 0:00:27 iter: 14879 total_loss: 0.135 loss_cls: 0.068 loss_box_reg: 0.064 time: 0.2293 data_time: 0.0142 lr: 0.000000 max_mem: 10044M [08/07 15:39:14 d2.utils.events]: eta: 0:00:23 iter: 14899 total_loss: 0.149 loss_cls: 0.077 loss_box_reg: 0.070 time: 0.2293 data_time: 0.0122 lr: 0.000000 max_mem: 10044M [08/07 15:39:18 d2.utils.events]: eta: 0:00:18 iter: 14919 total_loss: 0.140 loss_cls: 0.073 loss_box_reg: 0.067 time: 0.2293 data_time: 0.0141 lr: 0.000000 max_mem: 10044M [08/07 15:39:23 d2.utils.events]: eta: 0:00:13 iter: 14939 total_loss: 0.141 loss_cls: 0.075 loss_box_reg: 0.066 time: 0.2293 data_time: 0.0133 lr: 0.000000 max_mem: 10044M [08/07 15:39:28 d2.utils.events]: eta: 0:00:09 iter: 14959 total_loss: 0.139 loss_cls: 0.071 loss_box_reg: 0.068 time: 0.2293 data_time: 0.0123 lr: 0.000000 max_mem: 10044M [08/07 15:39:32 d2.utils.events]: eta: 0:00:04 iter: 14979 total_loss: 0.139 loss_cls: 0.077 loss_box_reg: 0.070 time: 0.2293 data_time: 0.0119 lr: 0.000000 max_mem: 10044M [08/07 15:39:42 d2.utils.events]: eta: 0:00:00 iter: 14999 total_loss: 0.140 loss_cls: 0.071 loss_box_reg: 0.067 time: 0.2293 data_time: 0.0113 lr: 0.000000 max_mem: 10044M [08/07 15:39:42 d2.engine.hooks]: Overall training speed: 9997 iterations in 0:38:12 (0.2293 s / it) [08/07 15:39:42 d2.engine.hooks]: Total training time: 0:38:28 (0:00:15 on hooks)
In [ ]:
def cfg_test():
cfg = get_cfg()
cfg.merge_from_file(model_zoo.get_config_file(MODEL_PATH))
cfg.MODEL.WEIGHTS = f'{OUTPUT_PATH}/model_final.pth'
cfg.DATASETS.TEST = ('wheat_test',)
cfg.MODEL.RETINANET.NUM_CLASSES = 1
cfg.MODEL.RETINANET.SCORE_THRESH_TEST = 0.4
return cfg
cfg = cfg_test()
predict = DefaultPredictor(cfg)
Results - Visualization¶
In [64]:
df_sub = pd.read_csv('/content/drive/My Drive/CV/Global Wheat Detection/sample_submission.csv')
TEST_DIR = '/content/drive/My Drive/CV/Global Wheat Detection/test'
fig, ax = plt.subplots(2, 5, figsize=(30, 17))
subplot_indexes = [(x,y) for x in range(2) for y in range(5)]
for index, image_id in enumerate(df_sub['image_id']):
im = cv2.imread('{}/{}.jpg'.format(TEST_DIR, image_id))
boxes = []
scores = []
labels = []
outputs = predict(im)
out = outputs["instances"].to("cpu")
scores = out.get_fields()['scores'].numpy()
boxes = out.get_fields()['pred_boxes'].tensor.numpy().astype(int)
labels= out.get_fields()['scores'].numpy()
boxes = boxes.astype(int)
boxes[:, 2] = boxes[:, 2] - boxes[:, 0]
boxes[:, 3] = boxes[:, 3] - boxes[:, 1]
im = cv2.cvtColor(im, cv2.COLOR_BGR2RGB).astype(np.float32)
im /= 255.0
for b,s in zip(boxes,scores):
cv2.rectangle(im, (b[0],b[1]), (b[0]+b[2],b[1]+b[3]), (1,1,1), 3)
cv2.putText(im, '{:.2}'.format(s), (b[0],b[1]), cv2.FONT_HERSHEY_COMPLEX_SMALL, 1, (1,1,1), 2)
ax[subplot_indexes[index]].set_axis_off()
ax[subplot_indexes[index]].imshow(im)
fig.tight_layout()
fig.show()
Conclusions¶
I got a low score on the leader boards with this model, I was very disappointed with the models low performance. I searched for performance boosting for the detectron2 but I haven't found any sources(maybe because its still new), I tried tuning the parameters, tried several models, but in the end the model's score wasn't getting any higher. I believe that I could boost its performance by adding more custom augmentations or change the LR scheduler, but because of lack of time I haven't done any extra work. On the future I will get the most of this model (maybe on different competition :)).
In [ ]: